{"version": 3, "context": "ComputeGraph { inputs: [GraphInput { name: \"vit.image.patches\", kind: Host, value: ValueId(0), shape: TensorShape([1, 4, 588]) }, GraphInput { name: \"vit.embedding.weight\", kind: Parameter, value: ValueId(1), shape: TensorShape([588, 144]) }, GraphInput { name: \"vit.embedding.bias\", kind: Parameter, value: ValueId(2), shape: TensorShape([1, 1, 144]) }, GraphInput { name: \"vit.position\", kind: Parameter, value: ValueId(5), shape: TensorShape([1, 4, 144]) }, GraphInput { name: \"vit.encoder.layer0.attention_norm.scale\", kind: Parameter, value: ValueId(7), shape: TensorShape([1, 1, 144]) }, GraphInput { name: \"vit.encoder.layer1.attention_norm.scale\", kind: Parameter, value: ValueId(8), shape: TensorShape([1, 1, 144]) }, GraphInput { name: \"vit.encoder.layer2.attention_norm.scale\", kind: Parameter, value: ValueId(9), shape: TensorShape([1, 1, 144]) }, GraphInput { name: \"vit.encoder.layer0.attention_norm.bias\", kind: Parameter, value: ValueId(10), shape: TensorShape([1, 1, 144]) }, GraphInput { name: \"vit.encoder.layer1.attention_norm.bias\", kind: Parameter, value: ValueId(11), shape: TensorShape([1, 1, 144]) }, GraphInput { name: \"vit.encoder.layer2.attention_norm.bias\", kind: Parameter, value: ValueId(12), shape: TensorShape([1, 1, 144]) }, GraphInput { name: \"vit.encoder.layer0.attention.qkv.weight\", kind: Parameter, value: ValueId(13), shape: TensorShape([144, 432]) }, GraphInput { name: \"vit.encoder.layer1.attention.qkv.weight\", kind: Parameter, value: ValueId(14), shape: TensorShape([144, 432]) }, GraphInput { name: \"vit.encoder.layer2.attention.qkv.weight\", kind: Parameter, value: ValueId(15), shape: TensorShape([144, 432]) }, GraphInput { name: \"vit.encoder.layer0.attention.qkv.bias\", kind: Parameter, value: ValueId(16), shape: TensorShape([1, 1, 432]) }, GraphInput { name: \"vit.encoder.layer1.attention.qkv.bias\", kind: Parameter, value: ValueId(17), shape: TensorShape([1, 1, 432]) }, GraphInput { name: \"vit.encoder.layer2.attention.qkv.bias\", kind: Parameter, value: ValueId(18), shape: TensorShape([1, 1, 432]) }, GraphInput { name: \"vit.encoder.layer0.attention.output.weight\", kind: Parameter, value: ValueId(19), shape: TensorShape([144, 144]) }, GraphInput { name: \"vit.encoder.layer1.attention.output.weight\", kind: Parameter, value: ValueId(20), shape: TensorShape([144, 144]) }, GraphInput { name: \"vit.encoder.layer2.attention.output.weight\", kind: Parameter, value: ValueId(21), shape: TensorShape([144, 144]) }, GraphInput { name: \"vit.encoder.layer0.attention.output.bias\", kind: Parameter, value: ValueId(22), shape: TensorShape([1, 1, 144]) }, GraphInput { name: \"vit.encoder.layer1.attention.output.bias\", kind: Parameter, value: ValueId(23), shape: TensorShape([1, 1, 144]) }, GraphInput { name: \"vit.encoder.layer2.attention.output.bias\", kind: Parameter, value: ValueId(24), shape: TensorShape([1, 1, 144]) }, GraphInput { name: \"vit.encoder.layer0.mlp_norm.scale\", kind: Parameter, value: ValueId(25), shape: TensorShape([1, 1, 144]) }, GraphInput { name: \"vit.encoder.layer1.mlp_norm.scale\", kind: Parameter, value: ValueId(26), shape: TensorShape([1, 1, 144]) }, GraphInput { name: \"vit.encoder.layer2.mlp_norm.scale\", kind: Parameter, value: ValueId(27), shape: TensorShape([1, 1, 144]) }, GraphInput { name: \"vit.encoder.layer0.mlp_norm.bias\", kind: Parameter, value: ValueId(28), shape: TensorShape([1, 1, 144]) }, GraphInput { name: \"vit.encoder.layer1.mlp_norm.bias\", kind: Parameter, value: ValueId(29), shape: TensorShape([1, 1, 144]) }, GraphInput { name: \"vit.encoder.layer2.mlp_norm.bias\", kind: Parameter, value: ValueId(30), shape: TensorShape([1, 1, 144]) }, GraphInput { name: \"vit.encoder.layer0.mlp.up.weight\", kind: Parameter, value: ValueId(31), shape: TensorShape([144, 288]) }, GraphInput { name: \"vit.encoder.layer1.mlp.up.weight\", kind: Parameter, value: ValueId(32), shape: TensorShape([144, 288]) }, GraphInput { name: \"vit.encoder.layer2.mlp.up.weight\", kind: Parameter, value: ValueId(33), shape: TensorShape([144, 288]) }, GraphInput { name: \"vit.encoder.layer0.mlp.up.bias\", kind: Parameter, value: ValueId(34), shape: TensorShape([1, 1, 288]) }, GraphInput { name: \"vit.encoder.layer1.mlp.up.bias\", kind: Parameter, value: ValueId(35), shape: TensorShape([1, 1, 288]) }, GraphInput { name: \"vit.encoder.layer2.mlp.up.bias\", kind: Parameter, value: ValueId(36), shape: TensorShape([1, 1, 288]) }, GraphInput { name: \"vit.encoder.layer0.mlp.down.weight\", kind: Parameter, value: ValueId(37), shape: TensorShape([288, 144]) }, GraphInput { name: \"vit.encoder.layer1.mlp.down.weight\", kind: Parameter, value: ValueId(38), shape: TensorShape([288, 144]) }, GraphInput { name: \"vit.encoder.layer2.mlp.down.weight\", kind: Parameter, value: ValueId(39), shape: TensorShape([288, 144]) }, GraphInput { name: \"vit.encoder.layer0.mlp.down.bias\", kind: Parameter, value: ValueId(40), shape: TensorShape([1, 1, 144]) }, GraphInput { name: \"vit.encoder.layer1.mlp.down.bias\", kind: Parameter, value: ValueId(41), shape: TensorShape([1, 1, 144]) }, GraphInput { name: \"vit.encoder.layer2.mlp.down.bias\", kind: Parameter, value: ValueId(42), shape: TensorShape([1, 1, 144]) }, GraphInput { name: \"vit.encoder.final_norm.scale\", kind: Parameter, value: ValueId(78), shape: TensorShape([1, 1, 144]) }, GraphInput { name: \"vit.encoder.final_norm.bias\", kind: Parameter, value: ValueId(79), shape: TensorShape([1, 1, 144]) }, GraphInput { name: \"vit.map.probe\", kind: Parameter, value: ValueId(81), shape: TensorShape([1, 1, 144]) }, GraphInput { name: \"vit.map.attention.kv.weight\", kind: Parameter, value: ValueId(82), shape: TensorShape([144, 288]) }, GraphInput { name: \"vit.map.attention.kv.bias\", kind: Parameter, value: ValueId(83), shape: TensorShape([1, 1, 288]) }, GraphInput { name: \"vit.map.attention.query.weight\", kind: Parameter, value: ValueId(86), shape: TensorShape([144, 144]) }, GraphInput { name: \"vit.map.attention.query.bias\", kind: Parameter, value: ValueId(87), shape: TensorShape([1, 1, 144]) }, GraphInput { name: \"vit.map.attention.output.weight\", kind: Parameter, value: ValueId(97), shape: TensorShape([144, 144]) }, GraphInput { name: \"vit.map.attention.output.bias\", kind: Parameter, value: ValueId(98), shape: TensorShape([1, 1, 144]) }, GraphInput { name: \"vit.map.norm.scale\", kind: Parameter, value: ValueId(101), shape: TensorShape([1, 1, 144]) }, GraphInput { name: \"vit.map.norm.bias\", kind: Parameter, value: ValueId(102), shape: TensorShape([1, 1, 144]) }, GraphInput { name: \"vit.map.mlp.up.weight\", kind: Parameter, value: ValueId(104), shape: TensorShape([144, 288]) }, GraphInput { name: \"vit.map.mlp.up.bias\", kind: Parameter, value: ValueId(105), shape: TensorShape([1, 1, 288]) }, GraphInput { name: \"vit.map.mlp.down.weight\", kind: Parameter, value: ValueId(109), shape: TensorShape([288, 144]) }, GraphInput { name: \"vit.map.mlp.down.bias\", kind: Parameter, value: ValueId(110), shape: TensorShape([1, 1, 144]) }], sequences: [ValueSequence { id: ValueSequenceId(0), name: \"vit.encoder.attention_norm.scale\", values: [ValueId(7), ValueId(8), ValueId(9)] }, ValueSequence { id: ValueSequenceId(1), name: \"vit.encoder.attention_norm.bias\", values: [ValueId(10), ValueId(11), ValueId(12)] }, ValueSequence { id: ValueSequenceId(2), name: \"vit.encoder.attention.qkv.weight\", values: [ValueId(13), ValueId(14), ValueId(15)] }, ValueSequence { id: ValueSequenceId(3), name: \"vit.encoder.attention.qkv.bias\", values: [ValueId(16), ValueId(17), ValueId(18)] }, ValueSequence { id: ValueSequenceId(4), name: \"vit.encoder.attention.output.weight\", values: [ValueId(19), ValueId(20), ValueId(21)] }, ValueSequence { id: ValueSequenceId(5), name: \"vit.encoder.attention.output.bias\", values: [ValueId(22), ValueId(23), ValueId(24)] }, ValueSequence { id: ValueSequenceId(6), name: \"vit.encoder.mlp_norm.scale\", values: [ValueId(25), ValueId(26), ValueId(27)] }, ValueSequence { id: ValueSequenceId(7), name: \"vit.encoder.mlp_norm.bias\", values: [ValueId(28), ValueId(29), ValueId(30)] }, ValueSequence { id: ValueSequenceId(8), name: \"vit.encoder.mlp.up.weight\", values: [ValueId(31), ValueId(32), ValueId(33)] }, ValueSequence { id: ValueSequenceId(9), name: \"vit.encoder.mlp.up.bias\", values: [ValueId(34), ValueId(35), ValueId(36)] }, ValueSequence { id: ValueSequenceId(10), name: \"vit.encoder.mlp.down.weight\", values: [ValueId(37), ValueId(38), ValueId(39)] }, ValueSequence { id: ValueSequenceId(11), name: \"vit.encoder.mlp.down.bias\", values: [ValueId(40), ValueId(41), ValueId(42)] }], 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(77)], kind: Repeat(Repeat { count: 3, 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(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)], operations: [Operation { id: OperationId(3), inputs: [ValueId(43), ValueId(44), ValueId(45)], results: [ValueId(56)], kind: LayerNorm }, Operation { id: OperationId(4), inputs: [ValueId(56), ValueId(46)], results: [ValueId(57)], kind: Gemm(GemmOptions { transpose_left: false, transpose_right: false }) }, Operation { id: OperationId(5), inputs: [ValueId(57), ValueId(47)], results: [ValueId(58)], kind: Add }, Operation { id: OperationId(6), inputs: [ValueId(58)], results: [ValueId(59)], kind: Slice(AxisSlice { axis: 2, start: 0, length: 144 }) }, Operation { id: OperationId(7), inputs: [ValueId(58)], results: [ValueId(60)], kind: Slice(AxisSlice { axis: 2, start: 144, length: 144 }) }, Operation { id: OperationId(8), inputs: [ValueId(58)], results: [ValueId(61)], kind: Slice(AxisSlice { axis: 2, start: 288, length: 144 }) }, Operation { id: OperationId(9), inputs: [ValueId(59)], results: [ValueId(62)], kind: View(AxisFactorView { split_axis: 2, merge_axis: 0, factor: 2, reversed: false }) }, Operation { id: OperationId(10), inputs: [ValueId(60)], results: [ValueId(63)], kind: View(AxisFactorView { split_axis: 2, merge_axis: 0, factor: 2, reversed: false }) }, Operation { id: OperationId(11), inputs: [ValueId(61)], results: [ValueId(64)], kind: View(AxisFactorView { split_axis: 2, merge_axis: 0, factor: 2, reversed: false }) }, Operation { id: OperationId(12), inputs: [ValueId(62), ValueId(63), ValueId(64)], results: [ValueId(65)], kind: FlashAttention(AttentionOptions { causal: false, scale: InverseSqrtQueryWidth }) }, Operation { id: OperationId(13), inputs: [ValueId(65)], results: [ValueId(66)], kind: View(AxisFactorView { split_axis: 2, merge_axis: 0, factor: 2, reversed: true }) }, Operation { id: OperationId(14), inputs: [ValueId(66), ValueId(48)], results: [ValueId(67)], kind: Gemm(GemmOptions { transpose_left: false, transpose_right: false }) }, Operation { id: OperationId(15), inputs: [ValueId(67), ValueId(49)], results: [ValueId(68)], kind: Add }, Operation { id: OperationId(16), inputs: [ValueId(43), ValueId(68)], results: [ValueId(69)], kind: Add }, Operation { id: OperationId(17), inputs: [ValueId(69), ValueId(50), ValueId(51)], results: [ValueId(70)], kind: LayerNorm }, Operation { id: OperationId(18), inputs: [ValueId(70), ValueId(52)], results: [ValueId(71)], kind: Gemm(GemmOptions { transpose_left: false, transpose_right: false }) }, Operation { id: OperationId(19), inputs: [ValueId(71), ValueId(53)], results: [ValueId(72)], kind: Add }, Operation { id: OperationId(20), inputs: [ValueId(72)], results: [ValueId(73)], kind: Gelu }, Operation { id: OperationId(21), inputs: [ValueId(73), ValueId(54)], results: [ValueId(74)], kind: Gemm(GemmOptions { transpose_left: false, transpose_right: false }) }, Operation { id: OperationId(22), inputs: [ValueId(74), ValueId(55)], results: [ValueId(75)], kind: Add }, Operation { id: OperationId(23), inputs: [ValueId(69), ValueId(75)], results: [ValueId(76)], kind: Add }], yields: [ValueId(76)], value_shapes: {ValueId(43): TensorShape([1, 4, 144]), ValueId(44): TensorShape([1, 1, 144]), ValueId(45): TensorShape([1, 1, 144]), ValueId(46): TensorShape([144, 432]), ValueId(47): TensorShape([1, 1, 432]), ValueId(48): TensorShape([144, 144]), ValueId(49): TensorShape([1, 1, 144]), ValueId(50): TensorShape([1, 1, 144]), ValueId(51): TensorShape([1, 1, 144]), ValueId(52): TensorShape([144, 288]), ValueId(53): TensorShape([1, 1, 288]), ValueId(54): TensorShape([288, 144]), ValueId(55): TensorShape([1, 1, 144]), ValueId(56): TensorShape([1, 4, 144]), ValueId(57): TensorShape([1, 4, 432]), ValueId(58): TensorShape([1, 4, 432]), ValueId(59): TensorShape([1, 4, 144]), ValueId(60): TensorShape([1, 4, 144]), ValueId(61): TensorShape([1, 4, 144]), ValueId(62): TensorShape([2, 4, 72]), ValueId(63): TensorShape([2, 4, 72]), ValueId(64): TensorShape([2, 4, 72]), ValueId(65): TensorShape([2, 4, 72]), ValueId(66): TensorShape([1, 4, 144]), ValueId(67): TensorShape([1, 4, 144]), ValueId(68): TensorShape([1, 4, 144]), ValueId(69): TensorShape([1, 4, 144]), ValueId(70): TensorShape([1, 4, 144]), ValueId(71): TensorShape([1, 4, 288]), ValueId(72): TensorShape([1, 4, 288]), ValueId(73): TensorShape([1, 4, 288]), ValueId(74): TensorShape([1, 4, 144]), ValueId(75): TensorShape([1, 4, 144]), ValueId(76): TensorShape([1, 4, 144])} } }) }, Operation { id: OperationId(25), inputs: [ValueId(77), ValueId(78), ValueId(79)], results: [ValueId(80)], kind: LayerNorm }, Operation { id: OperationId(26), inputs: [ValueId(80), ValueId(82)], results: [ValueId(84)], kind: Gemm(GemmOptions { transpose_left: false, transpose_right: false }) }, Operation { id: OperationId(27), inputs: [ValueId(84), ValueId(83)], results: [ValueId(85)], kind: Add }, Operation { id: OperationId(28), inputs: [ValueId(81), ValueId(86)], results: [ValueId(88)], kind: Gemm(GemmOptions { transpose_left: false, transpose_right: false }) }, Operation { id: OperationId(29), inputs: [ValueId(88), ValueId(87)], results: [ValueId(89)], kind: Add }, Operation { id: OperationId(30), inputs: [ValueId(85)], results: [ValueId(90)], kind: Slice(AxisSlice { axis: 2, start: 0, length: 144 }) }, Operation { id: OperationId(31), inputs: [ValueId(85)], results: [ValueId(91)], kind: Slice(AxisSlice { axis: 2, start: 144, length: 144 }) }, Operation { id: OperationId(32), inputs: [ValueId(89)], results: [ValueId(92)], kind: View(AxisFactorView { split_axis: 2, merge_axis: 0, factor: 2, reversed: false }) }, Operation { id: OperationId(33), inputs: [ValueId(90)], results: [ValueId(93)], kind: View(AxisFactorView { split_axis: 2, merge_axis: 0, factor: 2, reversed: false }) }, Operation { id: OperationId(34), inputs: [ValueId(91)], results: [ValueId(94)], kind: View(AxisFactorView { split_axis: 2, merge_axis: 0, factor: 2, reversed: false }) }, Operation { id: OperationId(35), inputs: [ValueId(92), ValueId(93), ValueId(94)], results: [ValueId(95)], kind: FlashAttention(AttentionOptions { causal: false, scale: InverseSqrtQueryWidth }) }, Operation { id: OperationId(36), inputs: [ValueId(95)], results: [ValueId(96)], kind: View(AxisFactorView { split_axis: 2, merge_axis: 0, factor: 2, reversed: true }) }, Operation { id: OperationId(37), inputs: [ValueId(96), ValueId(97)], results: [ValueId(99)], kind: Gemm(GemmOptions { transpose_left: false, transpose_right: false }) }, Operation { id: OperationId(38), inputs: [ValueId(99), ValueId(98)], results: [ValueId(100)], kind: Add }, Operation { id: OperationId(39), inputs: [ValueId(100), ValueId(101), ValueId(102)], results: [ValueId(103)], kind: LayerNorm }, Operation { id: OperationId(40), inputs: [ValueId(103), ValueId(104)], results: [ValueId(106)], kind: Gemm(GemmOptions { transpose_left: false, transpose_right: false }) }, Operation { id: OperationId(41), inputs: [ValueId(106), ValueId(105)], results: [ValueId(107)], kind: Add }, Operation { id: OperationId(42), inputs: [ValueId(107)], results: [ValueId(108)], kind: Gelu }, Operation { id: OperationId(43), inputs: [ValueId(108), ValueId(109)], results: [ValueId(111)], kind: Gemm(GemmOptions { transpose_left: false, transpose_right: false }) }, Operation { id: OperationId(44), inputs: [ValueId(111), ValueId(110)], results: [ValueId(112)], kind: Add }, Operation { id: OperationId(45), inputs: [ValueId(100), ValueId(112)], results: [ValueId(113)], kind: Add }], outputs: [ValueId(113)], shapes: {ValueId(0): TensorShape([1, 4, 588]), ValueId(1): TensorShape([588, 144]), ValueId(2): TensorShape([1, 1, 144]), ValueId(3): TensorShape([1, 4, 144]), ValueId(4): TensorShape([1, 4, 144]), ValueId(5): TensorShape([1, 4, 144]), ValueId(6): TensorShape([1, 4, 144]), ValueId(7): TensorShape([1, 1, 144]), ValueId(8): TensorShape([1, 1, 144]), ValueId(9): TensorShape([1, 1, 144]), ValueId(10): TensorShape([1, 1, 144]), ValueId(11): TensorShape([1, 1, 144]), ValueId(12): TensorShape([1, 1, 144]), ValueId(13): TensorShape([144, 432]), ValueId(14): TensorShape([144, 432]), ValueId(15): TensorShape([144, 432]), ValueId(16): TensorShape([1, 1, 432]), ValueId(17): TensorShape([1, 1, 432]), ValueId(18): TensorShape([1, 1, 432]), ValueId(19): TensorShape([144, 144]), ValueId(20): TensorShape([144, 144]), ValueId(21): TensorShape([144, 144]), ValueId(22): TensorShape([1, 1, 144]), ValueId(23): TensorShape([1, 1, 144]), ValueId(24): TensorShape([1, 1, 144]), ValueId(25): TensorShape([1, 1, 144]), ValueId(26): TensorShape([1, 1, 144]), ValueId(27): TensorShape([1, 1, 144]), ValueId(28): TensorShape([1, 1, 144]), ValueId(29): TensorShape([1, 1, 144]), ValueId(30): TensorShape([1, 1, 144]), ValueId(31): TensorShape([144, 288]), ValueId(32): TensorShape([144, 288]), ValueId(33): TensorShape([144, 288]), ValueId(34): TensorShape([1, 1, 288]), ValueId(35): TensorShape([1, 1, 288]), ValueId(36): TensorShape([1, 1, 288]), ValueId(37): TensorShape([288, 144]), ValueId(38): TensorShape([288, 144]), ValueId(39): TensorShape([288, 144]), ValueId(40): TensorShape([1, 1, 144]), ValueId(41): TensorShape([1, 1, 144]), ValueId(42): TensorShape([1, 1, 144]), ValueId(77): TensorShape([1, 4, 144]), ValueId(78): TensorShape([1, 1, 144]), ValueId(79): TensorShape([1, 1, 144]), ValueId(80): TensorShape([1, 4, 144]), ValueId(81): TensorShape([1, 1, 144]), ValueId(82): TensorShape([144, 288]), ValueId(83): TensorShape([1, 1, 288]), ValueId(84): TensorShape([1, 4, 288]), ValueId(85): TensorShape([1, 4, 288]), ValueId(86): TensorShape([144, 144]), ValueId(87): TensorShape([1, 1, 144]), ValueId(88): TensorShape([1, 1, 144]), ValueId(89): TensorShape([1, 1, 144]), ValueId(90): TensorShape([1, 4, 144]), ValueId(91): TensorShape([1, 4, 144]), ValueId(92): TensorShape([2, 1, 72]), ValueId(93): TensorShape([2, 4, 72]), ValueId(94): TensorShape([2, 4, 72]), ValueId(95): TensorShape([2, 1, 72]), ValueId(96): TensorShape([1, 1, 144]), ValueId(97): TensorShape([144, 144]), ValueId(98): TensorShape([1, 1, 144]), ValueId(99): TensorShape([1, 1, 144]), ValueId(100): TensorShape([1, 1, 144]), ValueId(101): TensorShape([1, 1, 144]), ValueId(102): TensorShape([1, 1, 144]), ValueId(103): TensorShape([1, 1, 144]), ValueId(104): TensorShape([144, 288]), ValueId(105): TensorShape([1, 1, 288]), ValueId(106): TensorShape([1, 1, 288]), ValueId(107): TensorShape([1, 1, 288]), ValueId(108): TensorShape([1, 1, 288]), ValueId(109): TensorShape([288, 144]), ValueId(110): TensorShape([1, 1, 144]), ValueId(111): TensorShape([1, 1, 144]), ValueId(112): TensorShape([1, 1, 144]), ValueId(113): TensorShape([1, 1, 144])}, next_operation: 46, next_value: 114 }\nPipelineConfig { tile_count: 64, optimization_steps: 0, load_search_state: None, save_search_state: None, capacity_baseline: false, 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): F8F143 { scale_exponent: -4 }, ValueId(14): F8F143 { scale_exponent: -4 }, ValueId(15): F8F143 { scale_exponent: -4 }, ValueId(16): F16, ValueId(17): F16, ValueId(18): F16, ValueId(19): F8F143 { scale_exponent: -4 }, ValueId(20): F8F143 { scale_exponent: -4 }, ValueId(21): F8F143 { scale_exponent: -4 }, 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): F8F143 { scale_exponent: -4 }, ValueId(32): F8F143 { scale_exponent: -4 }, ValueId(33): F8F143 { scale_exponent: -4 }, ValueId(34): F16, ValueId(35): F16, ValueId(36): F16, ValueId(37): F8F143 { scale_exponent: -4 }, ValueId(38): F8F143 { scale_exponent: -4 }, ValueId(39): F8F143 { scale_exponent: -4 }, ValueId(40): F16, ValueId(41): F16, ValueId(42): F16, ValueId(78): F16, ValueId(79): F16, ValueId(81): F16, ValueId(82): F8F143 { scale_exponent: -4 }, ValueId(83): F16, ValueId(86): F8F143 { scale_exponent: -4 }, ValueId(87): F16, ValueId(97): F8F143 { scale_exponent: -4 }, ValueId(98): F16, ValueId(101): F16, ValueId(102): F16, ValueId(104): F8F143 { scale_exponent: -4 }, ValueId(105): F16, ValueId(109): F8F143 { scale_exponent: -4 }, ValueId(110): 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: 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]) } }, 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]) } }, 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]) } }, 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 } }, 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 } }, 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]) } }, 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]) } }, 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]) } }, 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 } }, 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]) } }, 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]) } }, 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]) } }, 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 } }, 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 } }, 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]) } }, 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]) } }, 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]) } }, 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 } }, 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]) } }, 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]) } }, 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]) } }, 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 } }, 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 } }, 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]) } }, 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]) } }, 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]) } }, 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 } }, 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]) } }, 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]) } }, 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]) } }, 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 } }, 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 } }, 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]) } }, 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]) } }, 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]) } }, 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 } }, 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]) } }, 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]) } }, 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]) } }, 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 } }, 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 } }, 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]) } }, 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]) } }, 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]) } }, 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 } }, 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]) } }, 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]) } }, 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]) } }, 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 } }, 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 } }, 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]) } }, 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]) } }, 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]) } }, 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 } }, 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]) } }, 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]) } }, 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]) } }, 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 } }, 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 } }, 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]) } }, 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]) } }, 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]) } }, 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 } }, format_policy: Concrete }), ParallelGemm { tile_count: 64, 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: 81920, 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": 96, "output_column_block": 16, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "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": 63, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "1": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "2": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "3": {"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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 4, "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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "4": {"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": 27, "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": 54, "replicas": 27, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 54, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 54, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 27}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "5": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "6": {"operator": {"Slice": {"axis": 2, "start": 0, "length": 144}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "7": {"operator": {"Slice": {"axis": 2, "start": 144, "length": 144}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "8": {"operator": {"Slice": {"axis": 2, "start": 288, "length": 144}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "9": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "10": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "11": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "12": {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 4, "columns": 1, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "13": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "14": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "15": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "16": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, "17": {"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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 4, "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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "18": {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "19": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "20": {"operator": "Gelu", "dispatch": {"Pointwise": {"kernel": "Gelu"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "21": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "22": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "23": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, "25": {"operator": "LayerNorm", "dispatch": {"LayerNorm": {"parts": 4}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "26": {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "27": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "28": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "29": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "30": {"operator": {"Slice": {"axis": 2, "start": 0, "length": 144}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "31": {"operator": {"Slice": {"axis": 2, "start": 144, "length": 144}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "32": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "33": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "34": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "35": {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 1, "columns": 1, "inner": 1}, "probability_value": null, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "36": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "37": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "38": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "39": {"operator": "LayerNorm", "dispatch": {"LayerNorm": {"parts": 4}}, "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": 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"}}, "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": 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"}}, "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": 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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "40": {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "41": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "42": {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "43": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "44": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "45": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 36}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 36}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 36}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}}, "open_boundaries": [58, 69, 72], "cast_before_copies": [], "packing_rows": null, "parallel_reductions": 0, "disjoint_copy_sources": false, "cast_storage": {"default": "Separate", "operators": {}, "sites": []}}, "alternatives": {"0": [{"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": 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": 63, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 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": 63, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "inner_partitions": 7, "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": 63, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "inner_partitions": 7, "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": 63, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 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": 63, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 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": 63, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "inner_partitions": 7, "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": 63, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "inner_partitions": 7, "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": 63, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 3, "inner_partitions": 19, "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": 57, "replicas": 3, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 57, "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": 19, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 3, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 3}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 3, "inner_partitions": 19, "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": 57, "replicas": 3, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 57, "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": 19, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 3, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 3}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 3, "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": 57, "replicas": 3, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 57, "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": 19, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 3, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 3}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 3, "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": 57, "replicas": 3, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 57, "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": 19, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 3, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 3}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 3, "inner_partitions": 19, "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": 57, "replicas": 3, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 57, "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": 19, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 3, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 3}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 3, "inner_partitions": 19, "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": 57, "replicas": 3, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 57, "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": 19, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 3, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 3}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 3, "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": 57, "replicas": 3, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 57, "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": 19, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 3, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 3}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 3, "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": 57, "replicas": 3, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 57, "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": 19, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 3, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 3}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "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": 45, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "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": 45, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "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": 45, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "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": 45, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "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": 45, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "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": 45, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "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": 45, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "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": 45, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 6, "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": 60, "replicas": 6, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 6, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 6, "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": 60, "replicas": 6, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 6, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 6, "inner_partitions": 10, "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": 60, "replicas": 6, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 6, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 6, "inner_partitions": 10, "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": 60, "replicas": 6, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 6, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 6, "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": 60, "replicas": 6, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 6, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 6, "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": 60, "replicas": 6, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 6, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 6, "inner_partitions": 10, "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": 60, "replicas": 6, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 6, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 6, "inner_partitions": 10, "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": 60, "replicas": 6, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 6, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 1, "column_partitions": 9, "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": 36, "replicas": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 36, "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": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 1, "column_partitions": 9, "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": 36, "replicas": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 36, "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": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 1, "column_partitions": 9, "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": 36, "replicas": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 36, "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": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 1, "column_partitions": 9, "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": 36, "replicas": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 36, "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": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 1, "column_partitions": 9, "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": 36, "replicas": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 36, "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": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 1, "column_partitions": 9, "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": 36, "replicas": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 36, "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": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 1, "column_partitions": 9, "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": 36, "replicas": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 36, "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": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 1, "column_partitions": 9, "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": 36, "replicas": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 36, "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": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 3, "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": 60, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 6, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 3}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 3, "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": 60, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 6, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 3}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 3, "inner_partitions": 10, "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": 60, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 6, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 3}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 3, "inner_partitions": 10, "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": 60, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 6, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 3}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 3, "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": 60, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 6, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 3, "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": 60, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 6, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 3, "inner_partitions": 10, "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": 60, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 6, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 3, "inner_partitions": 10, "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": 60, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 6, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 3, "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": 60, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 6, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 3}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 3, "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": 60, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 6, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 3}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 3, "inner_partitions": 10, "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": 60, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 6, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 3}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 3, "inner_partitions": 10, "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": 60, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 6, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 3}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 3, "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": 60, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 6, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 3, "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": 60, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 6, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 3, "inner_partitions": 10, "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": 60, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 6, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 3, "inner_partitions": 10, "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": 60, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 6, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 9, "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": 54, "replicas": 9, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 224, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 3, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 9, "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": 54, "replicas": 9, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 224, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 3, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 9, "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": 54, "replicas": 9, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 224, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 3, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 9, "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": 54, "replicas": 9, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 224, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 3, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 9, "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": 54, "replicas": 9, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 224, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 3, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 9, "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": 54, "replicas": 9, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 224, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 3, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 9, "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": 54, "replicas": 9, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 224, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 3, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 9, "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": 54, "replicas": 9, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 224, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 3, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 9, "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": 54, "replicas": 9, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 224, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 3, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 9, "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": 54, "replicas": 9, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 224, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 3, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 9, "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": 54, "replicas": 9, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 224, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 3, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 9, "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": 54, "replicas": 9, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 224, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 3, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 9, "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": 54, "replicas": 9, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 224, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 3, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 9, "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": 54, "replicas": 9, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 224, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 3, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 9, "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": 54, "replicas": 9, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 224, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 3, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 9, "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": 54, "replicas": 9, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 224, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 3, "padding_groups": 1, "block_size": 224, "padding_multiple": 224, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 6, "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": 60, "replicas": 6, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 6, "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": 60, "replicas": 6, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 6, "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": 60, "replicas": 6, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 6, "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": 60, "replicas": 6, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 6, "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": 60, "replicas": 6, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 6, "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": 60, "replicas": 6, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 6, "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": 60, "replicas": 6, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 6, "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": 60, "replicas": 6, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 6, "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": 60, "replicas": 6, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 6, "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": 60, "replicas": 6, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 6, "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": 60, "replicas": 6, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 6, "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": 60, "replicas": 6, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 6, "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": 60, "replicas": 6, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 6, "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": 60, "replicas": 6, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 6, "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": 60, "replicas": 6, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 6, "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": 60, "replicas": 6, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 128, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 128, "padding_multiple": 128, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 8, "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": 64, "replicas": 8, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 16, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 8}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 8, "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": 64, "replicas": 8, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 16, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 8}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 8, "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": 64, "replicas": 8, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 16, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 8}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 8, "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": 64, "replicas": 8, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 16, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 8}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 8, "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": 64, "replicas": 8, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 16, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 8, "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": 64, "replicas": 8, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 16, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 8, "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": 64, "replicas": 8, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 16, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 8, "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": 64, "replicas": 8, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 16, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 8, "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": 64, "replicas": 8, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 16, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 8}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 8, "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": 64, "replicas": 8, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 16, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 8}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 8, "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": 64, "replicas": 8, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 16, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 8}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 8, "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": 64, "replicas": 8, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 16, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 8}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 8, "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": 64, "replicas": 8, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 16, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 8, "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": 64, "replicas": 8, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 16, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 8, "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": 64, "replicas": 8, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 16, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 8, "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": 64, "replicas": 8, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 160, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 4, "padding_groups": 1, "block_size": 160, "padding_multiple": 160, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 16, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}], "1": [{"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 4, "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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 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"}}, "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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "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"}}, "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": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}], "2": [{"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 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"}}, "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": 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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 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"}}, "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": 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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "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": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "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": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}], "3": [{"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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 4, "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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": "LayerNorm", "dispatch": {"LayerNorm": {"parts": 2}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": "LayerNorm", "dispatch": {"LayerNorm": {"parts": 4}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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"}}, "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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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"}}, "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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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"}}, "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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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"}}, "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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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"}}, "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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 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"}}, "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"}}, "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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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"}}, "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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "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"}}, "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"}}, "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": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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"}}, "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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}], "4": [{"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": 27, "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": 54, "replicas": 27, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 54, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 27, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 27}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 27, "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": 54, "replicas": 27, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 54, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 27, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 27}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 27, "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": 54, "replicas": 27, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 54, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 27, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 27}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 27, "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": 54, "replicas": 27, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 54, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 27, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 27}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 27, "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": 54, "replicas": 27, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 54, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 54, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 27}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 27, "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": 54, "replicas": 27, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 54, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 54, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 27}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 27, "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": 54, "replicas": 27, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 54, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 54, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 27}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 27, "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": 54, "replicas": 27, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 54, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 54, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 27}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 27, "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": 54, "replicas": 27, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 54, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 27, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 27}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 27, "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": 54, "replicas": 27, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 54, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 27, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 27}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 27, "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": 54, "replicas": 27, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 54, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 27, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 27}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 27, "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": 54, "replicas": 27, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 54, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 27, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 27}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 27, "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": 54, "replicas": 27, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 54, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 54, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 27}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 27, "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": 54, "replicas": 27, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 54, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 54, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 27}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 27, "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": 54, "replicas": 27, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 54, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 54, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 27}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 27, "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": 54, "replicas": 27, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 54, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 54, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 27}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 12, "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": 60, "replicas": 12, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 12}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 12, "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": 60, "replicas": 12, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 12}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 12, "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": 60, "replicas": 12, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 12}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 12, "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": 60, "replicas": 12, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 12}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 12, "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": 60, "replicas": 12, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 12}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 12, "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": 60, "replicas": 12, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 12}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 12, "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": 60, "replicas": 12, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 12}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 1, "column_partitions": 12, "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": 60, "replicas": 12, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 12}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 21, "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": 63, "replicas": 21, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 63, "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": 3, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 21, "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": 63, "replicas": 21, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 63, "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": 3, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 21, "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": 63, "replicas": 21, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 63, "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": 3, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 21, "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": 63, "replicas": 21, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 63, "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": 3, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 21, "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": 63, "replicas": 21, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 63, "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": 3, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 21, "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": 63, "replicas": 21, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 63, "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": 3, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 21, "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": 63, "replicas": 21, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 63, "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": 3, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 21, "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": 63, "replicas": 21, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 63, "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": 3, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 80, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 80, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 80, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 80, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 80, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 80, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 80, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 80, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 80, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 80, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 80, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 80, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 80, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 80, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 80, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 80, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 16, "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": 64, "replicas": 16, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 16}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 16, "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": 64, "replicas": 16, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 16}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 16, "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": 64, "replicas": 16, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 16}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 16, "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": 64, "replicas": 16, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 16}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 16, "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": 64, "replicas": 16, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 16, "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": 64, "replicas": 16, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 16, "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": 64, "replicas": 16, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 16, "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": 64, "replicas": 16, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 16, "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": 64, "replicas": 16, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 16}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 16, "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": 64, "replicas": 16, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 16}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 16, "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": 64, "replicas": 16, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 16}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 16, "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": 64, "replicas": 16, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 16}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 16, "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": 64, "replicas": 16, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 16, "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": 64, "replicas": 16, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 16, "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": 64, "replicas": 16, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 16, "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": 64, "replicas": 16, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 10, "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": 60, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 10, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 20, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 10, "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": 60, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 10, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 20, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 10, "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": 60, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 10, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 20, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 10, "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": 60, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 10, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 20, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 10, "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": 60, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 10, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 20, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 10, "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": 60, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 10, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 20, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 10, "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": 60, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 10, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 20, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 10, "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": 60, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 10, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 20, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 10, "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": 60, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 10, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 20, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 10, "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": 60, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 10, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 20, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 10, "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": 60, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 10, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 20, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 10, "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": 60, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 10, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 20, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 10, "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": 60, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 10, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 20, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 10, "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": 60, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 10, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 20, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 10, "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": 60, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 10, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 20, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 10, "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": 60, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 10, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 20, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 80, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 80, "padding_multiple": 80, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 80, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 80, "padding_multiple": 80, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 80, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 80, "padding_multiple": 80, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 80, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 80, "padding_multiple": 80, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 80, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 80, "padding_multiple": 80, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 80, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 80, "padding_multiple": 80, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 80, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 80, "padding_multiple": 80, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 80, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 80, "padding_multiple": 80, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 80, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 80, "padding_multiple": 80, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 80, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 80, "padding_multiple": 80, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 80, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 80, "padding_multiple": 80, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 80, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 80, "padding_multiple": 80, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 80, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 80, "padding_multiple": 80, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 80, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 80, "padding_multiple": 80, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 80, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 80, "padding_multiple": 80, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 80, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 6, "padding_groups": 1, "block_size": 80, "padding_multiple": 80, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}, {"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}], "5": [{"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 4, "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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 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"}}, "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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "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"}}, "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": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}], "6": [{"operator": {"Slice": {"axis": 2, "start": 0, "length": 144}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Slice": {"axis": 2, "start": 0, "length": 144}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}], "7": [{"operator": {"Slice": {"axis": 2, "start": 144, "length": 144}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Slice": {"axis": 2, "start": 144, "length": 144}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}], "8": [{"operator": {"Slice": {"axis": 2, "start": 288, "length": 144}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Slice": {"axis": 2, "start": 288, "length": 144}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}], "9": [{"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}], "10": [{"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}], "11": [{"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}], "12": [{"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": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 1, "columns": 1, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 1, "columns": 1, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 1, "columns": 1, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 1, "columns": 1, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 1, "columns": 1, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 1, "columns": 1, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 1, "columns": 1, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 1, "columns": 1, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 1, "columns": 1, "inner": 1}, "probability_value": null, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 1, "columns": 1, "inner": 1}, "probability_value": null, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 2, "columns": 1, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 2, "columns": 1, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 2, "columns": 1, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 2, "columns": 1, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 2, "columns": 1, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 2, "columns": 1, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 2, "columns": 1, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 2, "columns": 1, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 2, "columns": 1, "inner": 1}, "probability_value": null, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 2, "columns": 1, "inner": 1}, "probability_value": null, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 3, "columns": 1, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 3, "columns": 1, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 3, "columns": 1, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 3, "columns": 1, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 3, "columns": 1, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 3, "columns": 1, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 3, "columns": 1, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 3, "columns": 1, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 3, "columns": 1, "inner": 1}, "probability_value": null, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 3, "columns": 1, "inner": 1}, "probability_value": null, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 4, "columns": 1, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 4, "columns": 1, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 4, "columns": 1, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 4, "columns": 1, "inner": 1}, "probability_value": {"rows": 2, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 4, "columns": 1, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 4, "columns": 1, "inner": 1}, "probability_value": {"rows": 3, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 4, "columns": 1, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 4, "columns": 1, "inner": 1}, "probability_value": {"rows": 4, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 4, "columns": 1, "inner": 1}, "probability_value": null, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 4, "columns": 1, "inner": 1}, "probability_value": null, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": null, "probability_value": {"rows": 2, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": null, "probability_value": {"rows": 2, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": null, "probability_value": {"rows": 3, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": null, "probability_value": {"rows": 3, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": null, "probability_value": {"rows": 4, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": null, "probability_value": {"rows": 4, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}], "13": [{"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}], "14": [{"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "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": 54, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 3, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "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": 54, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 3, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "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": 54, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 3, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "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": 54, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 3, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "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": 54, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 3, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "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": 54, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 3, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "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": 54, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 3, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "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": 54, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 3, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "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": 54, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 3, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "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": 54, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 3, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "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": 54, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 3, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "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": 54, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 3, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "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": 54, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 3, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "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": 54, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 3, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "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": 54, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 3, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "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": 54, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 3, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}], "15": [{"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 4, "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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 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"}}, "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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "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"}}, "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": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}], "16": [{"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, {"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": 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"}}, "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": 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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, {"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": 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"}}, "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": 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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, {"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": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "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": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "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": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}], "17": [{"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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 4, "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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": "LayerNorm", "dispatch": {"LayerNorm": {"parts": 2}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": "LayerNorm", "dispatch": {"LayerNorm": {"parts": 4}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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"}}, "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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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"}}, "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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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"}}, "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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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"}}, "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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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"}}, "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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 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"}}, "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"}}, "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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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"}}, "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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "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"}}, "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"}}, "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": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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"}}, "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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}], "18": [{"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 12, "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": 60, "replicas": 12, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 12}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 12, "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": 60, "replicas": 12, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 12}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 12, "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": 60, "replicas": 12, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 12}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 12, "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": 60, "replicas": 12, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 12}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 12, "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": 60, "replicas": 12, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 12}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 12, "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": 60, "replicas": 12, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 12}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 12, "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": 60, "replicas": 12, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 12}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 12, "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": 60, "replicas": 12, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 12}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 36, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 36, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 36, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 36, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 36, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 36, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 36, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 36, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 36, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 36, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 36, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 36, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 36, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 36, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 36, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 36, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 36, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 36, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 36, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 36, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 36, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 36, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 36, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 36, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 36, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 36, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 36, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 36, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 36, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 36, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 36, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 36, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 36, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 36, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 36, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 36, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 36, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 36, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 36, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 36, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 10, "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": 60, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 10, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 20, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 10, "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": 60, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 10, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 20, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 10, "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": 60, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 10, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 20, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 10, "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": 60, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 10, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 20, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 10, "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": 60, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 10, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 20, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 10, "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": 60, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 10, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 20, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 10, "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": 60, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 10, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 20, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 10, "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": 60, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 10, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 20, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 10, "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": 60, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 10, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 20, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 10, "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": 60, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 10, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 20, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 10, "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": 60, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 10, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 20, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 10, "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": 60, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 10, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 20, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 10, "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": 60, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 10, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 20, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 10, "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": 60, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 10, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 20, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 10, "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": 60, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 10, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 20, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 10, "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": 60, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 10, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 20, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 16, "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": 64, "replicas": 16, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 16}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 16, "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": 64, "replicas": 16, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 16}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 16, "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": 64, "replicas": 16, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 16}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 16, "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": 64, "replicas": 16, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 16}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 16, "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": 64, "replicas": 16, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 16, "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": 64, "replicas": 16, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 16, "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": 64, "replicas": 16, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 16, "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": 64, "replicas": 16, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 16, "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": 64, "replicas": 16, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 16}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 16, "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": 64, "replicas": 16, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 16}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 16, "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": 64, "replicas": 16, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 16}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 16, "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": 64, "replicas": 16, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 16}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 16, "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": 64, "replicas": 16, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 16, "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": 64, "replicas": 16, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 16, "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": 64, "replicas": 16, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 16, "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": 64, "replicas": 16, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}], "19": [{"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 4, "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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 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"}}, "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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "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"}}, "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": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}], "20": [{"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"operator": "Gelu", "dispatch": {"Pointwise": {"kernel": "Gelu"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}], "21": [{"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 7, "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": 63, "replicas": 7, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 63, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 7, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 7, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 7, "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": 7}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 7, "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": 63, "replicas": 7, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 63, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 7, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 7, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 7, "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": 7}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 7, "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": 63, "replicas": 7, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 63, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 7, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 7, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 7, "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": 7}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 7, "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": 63, "replicas": 7, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 63, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 7, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 7, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 7, "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": 7}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 7, "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": 63, "replicas": 7, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 63, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 7, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 7, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 7, "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": 7}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 7, "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": 63, "replicas": 7, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 63, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 7, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 7, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 7, "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": 7}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 7, "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": 63, "replicas": 7, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 63, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 7, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 7, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 7, "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": 7}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 7, "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": 63, "replicas": 7, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 63, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 7, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 7, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 7, "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": 7}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 9, "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": 54, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 3, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 9, "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": 54, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 3, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 9, "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": 54, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 3, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 9, "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": 54, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 3, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 9, "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": 54, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 3, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 9, "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": 54, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 3, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 9, "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": 54, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 3, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 9, "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": 54, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 3, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 9, "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": 54, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 3, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 9, "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": 54, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 3, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 9, "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": 54, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 3, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 9, "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": 54, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 3, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 9, "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": 54, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 3, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 9, "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": 54, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 3, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 9, "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": 54, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 3, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 9, "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": 54, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 3, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 3, "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": 54, "replicas": 3, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 9, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 6, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 3}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 3, "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": 54, "replicas": 3, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 9, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 6, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 3}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 3, "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": 54, "replicas": 3, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 9, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 6, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 3}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 3, "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": 54, "replicas": 3, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 9, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 6, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 3}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 3, "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": 54, "replicas": 3, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 9, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 6, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 3, "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": 54, "replicas": 3, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 9, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 6, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 3, "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": 54, "replicas": 3, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 9, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 6, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 3, "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": 54, "replicas": 3, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 9, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 6, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 3, "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": 54, "replicas": 3, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 9, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 6, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 3}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 3, "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": 54, "replicas": 3, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 9, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 6, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 3}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 3, "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": 54, "replicas": 3, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 9, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 6, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 3}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 3, "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": 54, "replicas": 3, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 9, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 6, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 3}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 3, "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": 54, "replicas": 3, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 9, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 6, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 3, "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": 54, "replicas": 3, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 9, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 6, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 3, "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": 54, "replicas": 3, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 9, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 6, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 3, "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": 54, "replicas": 3, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 27, "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": 9, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 6, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "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": 6, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "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": 6, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "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": 6, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "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": 6, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "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": 6, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "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": 6, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "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": 6, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "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": 6, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "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": 6, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "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": 6, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "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": 6, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "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": 6, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "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": 6, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "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": 6, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "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": 6, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "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": 6, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}], "22": [{"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 4, "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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 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"}}, "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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "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"}}, "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": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}], "23": [{"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, {"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": 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"}}, "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": 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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, {"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": 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"}}, "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": 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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, {"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": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "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": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "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": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}], "25": [{"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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 4, "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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": "LayerNorm", "dispatch": {"LayerNorm": {"parts": 2}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": "LayerNorm", "dispatch": {"LayerNorm": {"parts": 4}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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"}}, "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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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"}}, "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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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"}}, "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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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"}}, "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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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"}}, "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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 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"}}, "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"}}, "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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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"}}, "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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "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"}}, "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"}}, "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": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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"}}, "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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}], "26": [{"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 12, "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": 60, "replicas": 12, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 12}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 12, "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": 60, "replicas": 12, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 12}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 12, "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": 60, "replicas": 12, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 12}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 12, "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": 60, "replicas": 12, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 12}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 12, "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": 60, "replicas": 12, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 12}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 12, "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": 60, "replicas": 12, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 12}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 12, "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": 60, "replicas": 12, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 12}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 12, "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": 60, "replicas": 12, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 12}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 6}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Gemm": {"options": {"transpose_left": false, "transpose_right": false}, "multiply": {"F8F143": {"scale_exponent": -4}}, "accumulate": "F16"}}, "dispatch": {"BlockedGemm": {"inner_block": 32, "output_column_block": 48, "orientation": "Normal", "distribution": {"ParallelReduction": {"row_partitions": 2, "column_partitions": 6, "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": 60, "replicas": 6, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 30, "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": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 36, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 36, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 36, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 36, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 36, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 36, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 36, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 36, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 36, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 36, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 36, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 36, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 36, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 36, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 36, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 36, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 36, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 36, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 36, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 36, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 36, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 36, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 36, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 36, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 36, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 36, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 36, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 36, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 36, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 36, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 36, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 36, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 36, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 36, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 36, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 36, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 36, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 36, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 36, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 36, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 10, "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": 60, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 10, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 20, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 10, "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": 60, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 10, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 20, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 10, "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": 60, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 10, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 20, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 10, "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": 60, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 10, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 20, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 10, "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": 60, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 10, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 20, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 10, "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": 60, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 10, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 20, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 10, "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": 60, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 10, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 20, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 10, "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": 60, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 10, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 20, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 10, "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": 60, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 10, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 20, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 10, "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": 60, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 10, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 20, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 10, "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": 60, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 10, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 20, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 10, "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": 60, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 10, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 20, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 10}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 10, "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": 60, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 10, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 20, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 10, "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": 60, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 10, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 20, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 10, "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": 60, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 10, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 20, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 10, "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": 60, "replicas": 10, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 10, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 20, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 16, "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": 64, "replicas": 16, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 16}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 16, "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": 64, "replicas": 16, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 16}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 16, "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": 64, "replicas": 16, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 16}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 16, "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": 64, "replicas": 16, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 16}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 16, "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": 64, "replicas": 16, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 16, "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": 64, "replicas": 16, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 16, "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": 64, "replicas": 16, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 16, "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": 64, "replicas": 16, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 16, "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": 64, "replicas": 16, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 16}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 16, "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": 64, "replicas": 16, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 16}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 16, "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": 64, "replicas": 16, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 16}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 16, "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": 64, "replicas": 16, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 16}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 16, "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": 64, "replicas": 16, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 16, "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": 64, "replicas": 16, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 16, "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": 64, "replicas": 16, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 2, "column_partitions": 16, "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": 64, "replicas": 16, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 32, "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": 2}, {"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}], "27": [{"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 32, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 4, "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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 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"}}, "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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "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"}}, "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": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}], "28": [{"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 12, "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": 32, "column_block": 16}}}, "tiling": {"tile_count": 60, "replicas": 12, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 60, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 12, "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": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 12, "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": 32, "column_block": 16}}}, "tiling": {"tile_count": 60, "replicas": 12, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 60, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 12, "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": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 12, "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": 32, "column_block": 16}}}, "tiling": {"tile_count": 60, "replicas": 12, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 60, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 12, "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": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 12, "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": 32, "column_block": 16}}}, "tiling": {"tile_count": 60, "replicas": 12, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 60, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 12, "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": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 12, "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": 32, "column_block": 16}}}, "tiling": {"tile_count": 60, "replicas": 12, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 60, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 12, "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": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 12, "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": 32, "column_block": 16}}}, "tiling": {"tile_count": 60, "replicas": 12, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 60, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 12, "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": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 12, "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": 32, "column_block": 16}}}, "tiling": {"tile_count": 60, "replicas": 12, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 60, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 12, "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": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 12, "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": 32, "column_block": 16}}}, "tiling": {"tile_count": 60, "replicas": 12, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 60, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 12, "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": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 12, "column_partitions": 1, "inner_partitions": 5, "result_row_partitions": 5, "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": 60, "replicas": 12, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 60, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 60, "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": 60, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 12, "column_partitions": 1, "inner_partitions": 5, "result_row_partitions": 5, "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": 60, "replicas": 12, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 60, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 60, "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": 60, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 12, "column_partitions": 1, "inner_partitions": 5, "result_row_partitions": 5, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 60, "replicas": 12, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 60, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 60, "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": 60, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 12, "column_partitions": 1, "inner_partitions": 5, "result_row_partitions": 5, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 60, "replicas": 12, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 60, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 60, "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": 60, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 12, "column_partitions": 1, "inner_partitions": 5, "result_row_partitions": 5, "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": 60, "replicas": 12, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 60, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 60, "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": 60, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 12, "column_partitions": 1, "inner_partitions": 5, "result_row_partitions": 5, "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": 60, "replicas": 12, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 60, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 60, "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": 60, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 12, "column_partitions": 1, "inner_partitions": 5, "result_row_partitions": 5, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 60, "replicas": 12, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 60, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 60, "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": 60, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 12, "column_partitions": 1, "inner_partitions": 5, "result_row_partitions": 5, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 60, "replicas": 12, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 60, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 60, "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": 60, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 12, "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": 32, "column_block": 16}}}, "tiling": {"tile_count": 60, "replicas": 12, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 60, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 12, "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": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 12, "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": 32, "column_block": 16}}}, "tiling": {"tile_count": 60, "replicas": 12, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 60, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 12, "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": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 12, "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": 32, "column_block": 16}}}, "tiling": {"tile_count": 60, "replicas": 12, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 60, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 12, "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": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 12, "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": 32, "column_block": 16}}}, "tiling": {"tile_count": 60, "replicas": 12, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 60, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 12, "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": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 12, "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": 32, "column_block": 16}}}, "tiling": {"tile_count": 60, "replicas": 12, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 60, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 12, "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": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 12, "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": 32, "column_block": 16}}}, "tiling": {"tile_count": 60, "replicas": 12, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 60, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 12, "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": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 12, "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": 32, "column_block": 16}}}, "tiling": {"tile_count": 60, "replicas": 12, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 60, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 12, "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": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 12, "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": 32, "column_block": 16}}}, "tiling": {"tile_count": 60, "replicas": 12, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 60, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 12, "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": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 12, "column_partitions": 1, "inner_partitions": 5, "result_row_partitions": 5, "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": 60, "replicas": 12, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 60, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 60, "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": 60, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 12, "column_partitions": 1, "inner_partitions": 5, "result_row_partitions": 5, "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": 60, "replicas": 12, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 60, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 60, "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": 60, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 12, "column_partitions": 1, "inner_partitions": 5, "result_row_partitions": 5, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 60, "replicas": 12, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 60, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 60, "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": 60, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 12, "column_partitions": 1, "inner_partitions": 5, "result_row_partitions": 5, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 60, "replicas": 12, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 60, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 60, "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": 60, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 12, "column_partitions": 1, "inner_partitions": 5, "result_row_partitions": 5, "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": 60, "replicas": 12, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 60, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 60, "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": 60, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 12, "column_partitions": 1, "inner_partitions": 5, "result_row_partitions": 5, "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": 60, "replicas": 12, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 60, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 60, "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": 60, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 12, "column_partitions": 1, "inner_partitions": 5, "result_row_partitions": 5, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 60, "replicas": 12, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 60, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 60, "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": 60, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 12, "column_partitions": 1, "inner_partitions": 5, "result_row_partitions": 5, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 60, "replicas": 12, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 60, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 12, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 60, "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": 60, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 21, "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": 64, "column_block": 16}}}, "tiling": {"tile_count": 63, "replicas": 21, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 63, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 21, "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": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 21, "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": 64, "column_block": 16}}}, "tiling": {"tile_count": 63, "replicas": 21, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 63, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 21, "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": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 21, "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": 64, "column_block": 16}}}, "tiling": {"tile_count": 63, "replicas": 21, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 63, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 21, "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": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 21, "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": 64, "column_block": 16}}}, "tiling": {"tile_count": 63, "replicas": 21, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 63, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 21, "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": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 21, "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": 64, "column_block": 16}}}, "tiling": {"tile_count": 63, "replicas": 21, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 63, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 21, "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": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 21, "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": 64, "column_block": 16}}}, "tiling": {"tile_count": 63, "replicas": 21, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 63, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 21, "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": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 21, "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": 64, "column_block": 16}}}, "tiling": {"tile_count": 63, "replicas": 21, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 63, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 21, "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": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 21, "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": 64, "column_block": 16}}}, "tiling": {"tile_count": 63, "replicas": 21, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 63, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 21, "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": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 21, "column_partitions": 1, "inner_partitions": 3, "result_row_partitions": 3, "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": 63, "replicas": 21, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 63, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 63, "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": 63, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 21, "column_partitions": 1, "inner_partitions": 3, "result_row_partitions": 3, "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": 63, "replicas": 21, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 63, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 63, "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": 63, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 21, "column_partitions": 1, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 63, "replicas": 21, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 63, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 63, "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": 63, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 21, "column_partitions": 1, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 63, "replicas": 21, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 63, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 63, "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": 63, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 21, "column_partitions": 1, "inner_partitions": 3, "result_row_partitions": 3, "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": 63, "replicas": 21, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 63, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 63, "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": 63, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 21, "column_partitions": 1, "inner_partitions": 3, "result_row_partitions": 3, "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": 63, "replicas": 21, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 63, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 63, "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": 63, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 21, "column_partitions": 1, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 63, "replicas": 21, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 63, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 63, "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": 63, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 21, "column_partitions": 1, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 63, "replicas": 21, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 63, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 63, "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": 63, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 21, "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": 64, "column_block": 16}}}, "tiling": {"tile_count": 63, "replicas": 21, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 63, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 21, "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": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 21, "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": 64, "column_block": 16}}}, "tiling": {"tile_count": 63, "replicas": 21, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 63, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 21, "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": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 21, "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": 64, "column_block": 16}}}, "tiling": {"tile_count": 63, "replicas": 21, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 63, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 21, "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": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 21, "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": 64, "column_block": 16}}}, "tiling": {"tile_count": 63, "replicas": 21, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 63, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 21, "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": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 21, "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": 64, "column_block": 16}}}, "tiling": {"tile_count": 63, "replicas": 21, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 63, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 21, "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": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 21, "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": 64, "column_block": 16}}}, "tiling": {"tile_count": 63, "replicas": 21, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 63, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 21, "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": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 21, "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": 64, "column_block": 16}}}, "tiling": {"tile_count": 63, "replicas": 21, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 63, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 21, "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": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 21, "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": 64, "column_block": 16}}}, "tiling": {"tile_count": 63, "replicas": 21, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 63, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 21, "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": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 21, "column_partitions": 1, "inner_partitions": 3, "result_row_partitions": 3, "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": 63, "replicas": 21, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 63, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 63, "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": 63, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 21, "column_partitions": 1, "inner_partitions": 3, "result_row_partitions": 3, "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": 63, "replicas": 21, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 63, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 63, "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": 63, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 21, "column_partitions": 1, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 63, "replicas": 21, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 63, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 63, "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": 63, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 21, "column_partitions": 1, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 63, "replicas": 21, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 63, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 63, "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": 63, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 21, "column_partitions": 1, "inner_partitions": 3, "result_row_partitions": 3, "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": 63, "replicas": 21, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 63, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 63, "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": 63, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 21, "column_partitions": 1, "inner_partitions": 3, "result_row_partitions": 3, "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": 63, "replicas": 21, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 63, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 63, "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": 63, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 21, "column_partitions": 1, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 63, "replicas": 21, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 63, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 63, "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": 63, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 21, "column_partitions": 1, "inner_partitions": 3, "result_row_partitions": 3, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 63, "replicas": 21, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 63, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 21, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 63, "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": 63, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 32, "column_partitions": 1, "inner_partitions": 2, "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": 64, "replicas": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 32, "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": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 32, "column_partitions": 1, "inner_partitions": 2, "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": 64, "replicas": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 32, "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": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 32, "column_partitions": 1, "inner_partitions": 2, "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": 96, "column_block": 16}}}, "tiling": {"tile_count": 64, "replicas": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 32, "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": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 32, "column_partitions": 1, "inner_partitions": 2, "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": 96, "column_block": 16}}}, "tiling": {"tile_count": 64, "replicas": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 32, "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": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 32, "column_partitions": 1, "inner_partitions": 2, "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": 64, "replicas": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 32, "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": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 32, "column_partitions": 1, "inner_partitions": 2, "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": 64, "replicas": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 32, "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": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 32, "column_partitions": 1, "inner_partitions": 2, "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": 96, "column_block": 16}}}, "tiling": {"tile_count": 64, "replicas": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 32, "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": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 32, "column_partitions": 1, "inner_partitions": 2, "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": 96, "column_block": 16}}}, "tiling": {"tile_count": 64, "replicas": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 32, "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": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 32, "column_partitions": 1, "inner_partitions": 2, "result_row_partitions": 2, "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": 64, "replicas": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 64, "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": 64, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 32, "column_partitions": 1, "inner_partitions": 2, "result_row_partitions": 2, "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": 64, "replicas": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 64, "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": 64, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 32, "column_partitions": 1, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 64, "replicas": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 64, "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": 64, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 32, "column_partitions": 1, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 64, "replicas": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 64, "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": 64, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 32, "column_partitions": 1, "inner_partitions": 2, "result_row_partitions": 2, "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": 64, "replicas": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 64, "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": 64, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 32, "column_partitions": 1, "inner_partitions": 2, "result_row_partitions": 2, "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": 64, "replicas": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 64, "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": 64, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 32, "column_partitions": 1, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 64, "replicas": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 64, "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": 64, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 32, "column_partitions": 1, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 64, "replicas": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 64, "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": 64, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 32, "column_partitions": 1, "inner_partitions": 2, "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": 64, "replicas": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 32, "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": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 32, "column_partitions": 1, "inner_partitions": 2, "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": 64, "replicas": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 32, "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": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 32, "column_partitions": 1, "inner_partitions": 2, "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": 96, "column_block": 16}}}, "tiling": {"tile_count": 64, "replicas": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 32, "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": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 32, "column_partitions": 1, "inner_partitions": 2, "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": 96, "column_block": 16}}}, "tiling": {"tile_count": 64, "replicas": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 32, "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": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 32, "column_partitions": 1, "inner_partitions": 2, "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": 64, "replicas": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 32, "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": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 32, "column_partitions": 1, "inner_partitions": 2, "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": 64, "replicas": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 32, "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": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 32, "column_partitions": 1, "inner_partitions": 2, "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": 96, "column_block": 16}}}, "tiling": {"tile_count": 64, "replicas": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 32, "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": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 32, "column_partitions": 1, "inner_partitions": 2, "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": 96, "column_block": 16}}}, "tiling": {"tile_count": 64, "replicas": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 32, "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": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 32, "column_partitions": 1, "inner_partitions": 2, "result_row_partitions": 2, "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": 64, "replicas": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 64, "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": 64, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 32, "column_partitions": 1, "inner_partitions": 2, "result_row_partitions": 2, "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": 64, "replicas": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 64, "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": 64, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 32, "column_partitions": 1, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 64, "replicas": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 64, "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": 64, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 32, "column_partitions": 1, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 64, "replicas": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 64, "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": 64, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 32, "column_partitions": 1, "inner_partitions": 2, "result_row_partitions": 2, "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": 64, "replicas": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 64, "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": 64, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 32, "column_partitions": 1, "inner_partitions": 2, "result_row_partitions": 2, "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": 64, "replicas": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 64, "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": 64, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 32, "column_partitions": 1, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 64, "replicas": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 64, "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": 64, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 32, "column_partitions": 1, "inner_partitions": 2, "result_row_partitions": 2, "result_column_partitions": 1, "reduction_staging": "Streamed"}}}}, "requirements": {"inputs": [{"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"TransposedMatrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 64, "replicas": 32, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "MatchRemote", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 64, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 32, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 64, "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": 64, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 1}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}], "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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 36}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 36}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 36}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 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"}}, "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": 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"}}, "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": 1}, "partitions": 32, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 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"}}, "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": 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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 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"}}, "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": 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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 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"}}, "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": 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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 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"}}, "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": 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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "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": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "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": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}], "30": [{"operator": {"Slice": {"axis": 2, "start": 0, "length": 144}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Slice": {"axis": 2, "start": 0, "length": 144}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}], "31": [{"operator": {"Slice": {"axis": 2, "start": 144, "length": 144}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"Slice": {"axis": 2, "start": 144, "length": 144}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}], "32": [{"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}], "33": [{"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}], "34": [{"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}], "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": 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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 1, "columns": 1, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 1, "columns": 1, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 1, "columns": 1, "inner": 1}, "probability_value": null, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 1, "columns": 1, "inner": 1}, "probability_value": null, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}], "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}], "36": [{"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}], "37": [{"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}], "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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 36}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 36}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 36}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 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"}}, "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": 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"}}, "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": 1}, "partitions": 32, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 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"}}, "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": 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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 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"}}, "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": 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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 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"}}, "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": 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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 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"}}, "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": 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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "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": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "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": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}], "39": [{"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": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "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": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "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": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "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": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": "LayerNorm", "dispatch": {"LayerNorm": {"parts": 2}}, "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": 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"}}, "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": 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"}}, "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": 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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"operator": "LayerNorm", "dispatch": {"LayerNorm": {"parts": 4}}, "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": 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"}}, "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": 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"}}, "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": 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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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"}}, "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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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"}}, "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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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"}}, "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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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"}}, "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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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"}}, "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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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"}}, "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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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"}}, "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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}], "40": [{"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 12, "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": 60, "replicas": 12, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 12}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 12, "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": 60, "replicas": 12, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 12}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 12, "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": 60, "replicas": 12, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 12}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 12, "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": 60, "replicas": 12, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 12}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 12, "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": 60, "replicas": 12, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 12}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 12, "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": 60, "replicas": 12, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 12}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 12, "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": 60, "replicas": 12, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 12}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 12, "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": 60, "replicas": 12, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 12, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 12}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 36, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 36, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 36, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 36, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 36, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 36, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 36, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 36, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 36, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 36, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 36, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 36, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 36, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 36, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 18, "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": 36, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 36, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}], "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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 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"}}, "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": 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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 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"}}, "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": 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"}}, "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": 1}, "partitions": 32, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 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"}}, "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": 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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 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"}}, "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": 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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 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"}}, "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": 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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 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"}}, "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": 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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "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": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "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": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}], "42": [{"operator": "Gelu", "dispatch": {"Pointwise": {"kernel": "Gelu"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 1, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}], "43": [{"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 7, "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": 63, "replicas": 7, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 63, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 7, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 7, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 7, "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": 7}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 7, "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": 63, "replicas": 7, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 63, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 7, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 7, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 7, "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": 7}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 7, "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": 63, "replicas": 7, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 63, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 7, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 7, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 7, "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": 7}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 7, "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": 63, "replicas": 7, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 63, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 7, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 7, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 7, "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": 7}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 7, "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": 63, "replicas": 7, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 63, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 7, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 7, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 7, "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": 7}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 7, "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": 63, "replicas": 7, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 63, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 7, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 7, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 7, "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": 7}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 7, "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": 63, "replicas": 7, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 63, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 7, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 7, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 7, "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": 7}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, {"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": 7, "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": 63, "replicas": 7, "axes": [{"axis": {"FromEnd": 1}, "partitions": 9, "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"}}, "local_staging": "Direct", "materialization": "DispatchSlices"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "column_block": 16}}}, "tiling": {"tile_count": 63, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 7, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 7, "replicas": 1, "axes": [{"axis": {"FromEnd": 1}, "partitions": 7, "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": 7}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}], "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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 36}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 36}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 36}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 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"}}, "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": 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"}}, "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": 1}, "partitions": 32, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 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"}}, "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": 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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 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"}}, "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": 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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 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"}}, "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": 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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 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"}}, "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": 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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, {"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": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "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": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "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": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}], "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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 36}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 36}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 36}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, {"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": 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"}}, "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": 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"}}, "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": 1}, "partitions": 32, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, {"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": 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"}}, "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": 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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, {"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": 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"}}, "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": 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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, {"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": 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"}}, "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": 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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, {"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": 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"}}, "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": 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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, {"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"}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, {"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": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "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": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "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": 1}, "partitions": 1, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}]}, "inputs": {"0": {"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"}}, "1": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 57, "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": 3, "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": 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"}}, "5": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 5, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 5, "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": 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"}}, "8": {"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"}}, "9": {"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"}}, "10": {"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"}}, "11": {"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"}}, "12": {"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"}}, "13": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 54, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 27, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "14": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 54, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 27, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "15": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 54, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 2, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 27, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "16": {"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"}}, "17": {"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"}}, "18": {"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"}}, "19": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 45, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "20": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 45, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "21": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 45, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "22": {"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"}}, "23": {"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"}}, "24": {"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"}}, "25": {"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"}}, "26": {"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"}}, "27": {"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"}}, "28": {"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"}}, "29": {"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"}}, "30": {"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"}}, "31": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 54, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "32": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 54, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "33": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 54, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "34": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 3, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 3, "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": 3, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 3, "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": 3, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 3, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "37": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 45, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "38": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 45, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "39": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 45, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "40": {"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"}}, "41": {"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"}}, "42": {"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"}}, "78": {"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"}}, "79": {"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"}}, "81": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 9, "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": 54, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "83": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 3, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 3, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "86": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 45, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "87": {"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"}}, "97": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 45, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "98": {"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"}}, "101": {"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"}}, "102": {"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"}}, "104": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 54, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 3, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 18, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "105": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 3, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 3, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "109": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"Amp": "TransposedLeft"}, "tiling": {"tile_count": 45, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 5, "padding_groups": 1, "block_size": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}, {"axis": {"FromEnd": 1}, "partitions": 9, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "110": {"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"}}}, "visited": [{"plans": {"0": {"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": 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": 63, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "1": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "2": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "3": {"operator": "LayerNorm", "dispatch": {"LayerNorm": {"parts": 4}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "4": {"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": 27, "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": 54, "replicas": 27, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 54, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 54, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 27}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "5": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "6": {"operator": {"Slice": {"axis": 2, "start": 0, "length": 144}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "7": {"operator": {"Slice": {"axis": 2, "start": 144, "length": 144}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "8": {"operator": {"Slice": {"axis": 2, "start": 288, "length": 144}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "9": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "10": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "11": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "12": {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 4, "columns": 1, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "13": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "14": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "15": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "16": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, "18": {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "19": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "20": {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "21": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "22": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "25": {"operator": "LayerNorm", "dispatch": {"LayerNorm": {"parts": 4}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "26": {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "27": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "28": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "29": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "30": {"operator": {"Slice": {"axis": 2, "start": 0, "length": 144}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "31": {"operator": {"Slice": {"axis": 2, "start": 144, "length": 144}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "32": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "33": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "34": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "35": {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 1, "columns": 1, "inner": 1}, "probability_value": null, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "36": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "37": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "38": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "39": {"operator": "LayerNorm", "dispatch": {"LayerNorm": {"parts": 4}}, "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": 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"}}, "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": 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"}}, "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": 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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "40": {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "41": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "42": {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "43": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "44": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "45": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 36}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 36}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 36}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}}, "open_boundaries": [69], "cast_before_copies": [], "packing_rows": null, "parallel_reductions": 0, "disjoint_copy_sources": false, "cast_storage": {"default": "Separate", "operators": {}, "sites": []}}, {"plans": {"0": {"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": 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": 63, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "1": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "2": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "3": {"operator": "LayerNorm", "dispatch": {"LayerNorm": {"parts": 4}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "4": {"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": 27, "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": 54, "replicas": 27, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 54, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 54, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 27}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "5": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "6": {"operator": {"Slice": {"axis": 2, "start": 0, "length": 144}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "7": {"operator": {"Slice": {"axis": 2, "start": 144, "length": 144}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "8": {"operator": {"Slice": {"axis": 2, "start": 288, "length": 144}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "9": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "10": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "11": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "12": {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 4, "columns": 1, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "13": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "14": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "15": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "16": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, "17": {"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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 4, "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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "18": {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "19": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "20": {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "21": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "22": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "23": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, "25": {"operator": "LayerNorm", "dispatch": {"LayerNorm": {"parts": 4}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "26": {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "27": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "28": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "29": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "30": {"operator": {"Slice": {"axis": 2, "start": 0, "length": 144}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "31": {"operator": {"Slice": {"axis": 2, "start": 144, "length": 144}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "32": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "33": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "34": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "35": {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 1, "columns": 1, "inner": 1}, "probability_value": null, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "36": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "37": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "38": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "39": {"operator": "LayerNorm", "dispatch": {"LayerNorm": {"parts": 4}}, "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": 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"}}, "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": 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"}}, "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": 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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "40": {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "41": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "42": {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "43": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "44": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "45": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 36}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 36}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 36}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}}, "open_boundaries": [69], "cast_before_copies": [], "packing_rows": null, "parallel_reductions": 0, "disjoint_copy_sources": false, "cast_storage": {"default": "Separate", "operators": {}, "sites": []}}, {"plans": {"0": {"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": 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": 63, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "1": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "2": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "3": {"operator": "LayerNorm", "dispatch": {"LayerNorm": {"parts": 4}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "4": {"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": 27, "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": 54, "replicas": 27, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 54, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 54, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 27}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "5": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "6": {"operator": {"Slice": {"axis": 2, "start": 0, "length": 144}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "7": {"operator": {"Slice": {"axis": 2, "start": 144, "length": 144}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "8": {"operator": {"Slice": {"axis": 2, "start": 288, "length": 144}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "9": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "10": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "11": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "12": {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 4, "columns": 1, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "13": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "14": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "15": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "16": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, "18": {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "19": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "20": {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "21": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "22": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "25": {"operator": "LayerNorm", "dispatch": {"LayerNorm": {"parts": 4}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "26": {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "27": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "28": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "29": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "30": {"operator": {"Slice": {"axis": 2, "start": 0, "length": 144}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "31": {"operator": {"Slice": {"axis": 2, "start": 144, "length": 144}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "32": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "33": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "34": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "35": {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 1, "columns": 1, "inner": 1}, "probability_value": null, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "36": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "37": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "38": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "39": {"operator": "LayerNorm", "dispatch": {"LayerNorm": {"parts": 4}}, "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": 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"}}, "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": 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"}}, "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": 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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "40": {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "41": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "42": {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "43": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "44": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "45": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 36}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 36}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 36}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}}, "open_boundaries": [69], "cast_before_copies": [], "packing_rows": null, "parallel_reductions": 0, "disjoint_copy_sources": false, "cast_storage": {"default": "Separate", "operators": {"16": "ReuseIfSmaller", "17": "ReuseIfSmaller", "23": "ReuseIfSmaller"}, "sites": []}}, {"plans": {"0": {"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": 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": 63, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "1": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "2": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "3": {"operator": "LayerNorm", "dispatch": {"LayerNorm": {"parts": 4}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "4": {"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": 27, "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": 54, "replicas": 27, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 54, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 54, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 27}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "5": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "6": {"operator": {"Slice": {"axis": 2, "start": 0, "length": 144}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "7": {"operator": {"Slice": {"axis": 2, "start": 144, "length": 144}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "8": {"operator": {"Slice": {"axis": 2, "start": 288, "length": 144}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "9": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "10": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "11": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "12": {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 4, "columns": 1, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "13": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "14": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "15": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "16": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, "17": {"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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 4, "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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "18": {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "19": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "20": {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "21": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "22": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "23": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, "25": {"operator": "LayerNorm", "dispatch": {"LayerNorm": {"parts": 4}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "26": {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "27": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "28": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "29": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "30": {"operator": {"Slice": {"axis": 2, "start": 0, "length": 144}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "31": {"operator": {"Slice": {"axis": 2, "start": 144, "length": 144}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "32": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "33": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "34": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "35": {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 1, "columns": 1, "inner": 1}, "probability_value": null, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "36": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "37": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "38": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "39": {"operator": "LayerNorm", "dispatch": {"LayerNorm": {"parts": 4}}, "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": 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"}}, "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": 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"}}, "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": 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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "40": {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "41": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "42": {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "43": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "44": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "45": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 36}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 36}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 36}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}}, "open_boundaries": [69], "cast_before_copies": [], "packing_rows": null, "parallel_reductions": 0, "disjoint_copy_sources": false, "cast_storage": {"default": "Separate", "operators": {"16": "ReuseIfSmaller", "17": "ReuseIfSmaller", "23": "ReuseIfSmaller"}, "sites": []}}, {"plans": {"0": {"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": 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": 63, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "1": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "2": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "3": {"operator": "LayerNorm", "dispatch": {"LayerNorm": {"parts": 4}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "4": {"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": 27, "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": 54, "replicas": 27, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 54, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 54, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 27}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "5": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "9": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "10": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "11": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "12": {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 4, "columns": 1, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "13": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "14": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "15": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "16": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, "17": {"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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 4, "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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "18": {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "19": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "20": {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "21": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "22": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "23": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, "25": {"operator": "LayerNorm", "dispatch": {"LayerNorm": {"parts": 4}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "26": {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "27": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "28": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "29": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "30": {"operator": {"Slice": {"axis": 2, "start": 0, "length": 144}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "31": {"operator": {"Slice": {"axis": 2, "start": 144, "length": 144}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "32": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "33": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "34": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "35": {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 1, "columns": 1, "inner": 1}, "probability_value": null, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "36": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "37": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "38": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "39": {"operator": "LayerNorm", "dispatch": {"LayerNorm": {"parts": 4}}, "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": 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"}}, "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": 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"}}, "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": 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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "40": {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "41": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "42": {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "43": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "44": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "45": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 36}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 36}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 36}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}}, "open_boundaries": [58, 69], "cast_before_copies": [], "packing_rows": null, "parallel_reductions": 0, "disjoint_copy_sources": false, "cast_storage": {"default": "Separate", "operators": {}, "sites": []}}, {"plans": {"0": {"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": 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": 63, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "1": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "2": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "3": {"operator": "LayerNorm", "dispatch": {"LayerNorm": {"parts": 4}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "4": {"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": 27, "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": 54, "replicas": 27, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 54, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 54, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 27}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "5": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "6": {"operator": {"Slice": {"axis": 2, "start": 0, "length": 144}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "7": {"operator": {"Slice": {"axis": 2, "start": 144, "length": 144}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "8": {"operator": {"Slice": {"axis": 2, "start": 288, "length": 144}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "9": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "10": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "11": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "12": {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 4, "columns": 1, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "13": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "14": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "15": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "16": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, "17": {"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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 4, "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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "18": {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "19": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "20": {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "21": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "22": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "23": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, "25": {"operator": "LayerNorm", "dispatch": {"LayerNorm": {"parts": 4}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "26": {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "27": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "28": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "29": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "30": {"operator": {"Slice": {"axis": 2, "start": 0, "length": 144}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "31": {"operator": {"Slice": {"axis": 2, "start": 144, "length": 144}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "32": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "33": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "34": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "35": {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 1, "columns": 1, "inner": 1}, "probability_value": null, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "36": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "37": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "38": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "39": {"operator": "LayerNorm", "dispatch": {"LayerNorm": {"parts": 4}}, "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": 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"}}, "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": 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"}}, "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": 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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "40": {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "41": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "42": {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "43": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "44": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "45": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 36}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 36}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 36}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}}, "open_boundaries": [58, 69], "cast_before_copies": [], "packing_rows": null, "parallel_reductions": 0, "disjoint_copy_sources": false, "cast_storage": {"default": "Separate", "operators": {}, "sites": []}}, {"plans": {"0": {"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": 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": 63, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "1": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "2": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "3": {"operator": "LayerNorm", "dispatch": {"LayerNorm": {"parts": 4}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "4": {"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": 27, "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": 54, "replicas": 27, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 54, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 54, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 27}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "5": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "9": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "10": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "11": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "12": {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 4, "columns": 1, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "13": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "14": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "15": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "16": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, "17": {"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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 4, "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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "18": {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "19": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "20": {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "21": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "22": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "23": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, "25": {"operator": "LayerNorm", "dispatch": {"LayerNorm": {"parts": 4}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "26": {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "27": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "28": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "29": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "30": {"operator": {"Slice": {"axis": 2, "start": 0, "length": 144}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "31": {"operator": {"Slice": {"axis": 2, "start": 144, "length": 144}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "32": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "33": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "34": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "35": {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 1, "columns": 1, "inner": 1}, "probability_value": null, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "36": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "37": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "38": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "39": {"operator": "LayerNorm", "dispatch": {"LayerNorm": {"parts": 4}}, "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": 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"}}, "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": 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"}}, "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": 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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "40": {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "41": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "42": {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "43": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "44": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "45": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 36}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 36}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 36}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}}, "open_boundaries": [58, 69], "cast_before_copies": [], "packing_rows": null, "parallel_reductions": 0, "disjoint_copy_sources": false, "cast_storage": {"default": "Separate", "operators": {"5": "ReuseIfSmaller", "6": "ReuseIfSmaller", "7": "ReuseIfSmaller", "8": "ReuseIfSmaller"}, "sites": []}}, {"plans": {"0": {"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": 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": 63, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "1": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "2": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "3": {"operator": "LayerNorm", "dispatch": {"LayerNorm": {"parts": 4}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "4": {"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": 27, "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": 54, "replicas": 27, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 54, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 54, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 27}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "5": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "6": {"operator": {"Slice": {"axis": 2, "start": 0, "length": 144}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "7": {"operator": {"Slice": {"axis": 2, "start": 144, "length": 144}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "8": {"operator": {"Slice": {"axis": 2, "start": 288, "length": 144}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "9": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "10": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "11": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "12": {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 4, "columns": 1, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "13": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "14": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "15": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "16": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, "17": {"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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 4, "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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "18": {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "19": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "20": {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "21": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "22": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "23": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, "25": {"operator": "LayerNorm", "dispatch": {"LayerNorm": {"parts": 4}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "26": {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "27": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "28": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "29": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "30": {"operator": {"Slice": {"axis": 2, "start": 0, "length": 144}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "31": {"operator": {"Slice": {"axis": 2, "start": 144, "length": 144}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "32": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "33": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "34": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "35": {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 1, "columns": 1, "inner": 1}, "probability_value": null, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "36": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "37": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "38": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "39": {"operator": "LayerNorm", "dispatch": {"LayerNorm": {"parts": 4}}, "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": 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"}}, "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": 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"}}, "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": 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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "40": {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "41": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "42": {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "43": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "44": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "45": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 36}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 36}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 36}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}}, "open_boundaries": [58, 69], "cast_before_copies": [], "packing_rows": null, "parallel_reductions": 0, "disjoint_copy_sources": false, "cast_storage": {"default": "Separate", "operators": {"5": "ReuseIfSmaller", "6": "ReuseIfSmaller", "7": "ReuseIfSmaller", "8": "ReuseIfSmaller"}, "sites": []}}, {"plans": {"0": {"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": 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": 63, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "1": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "2": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "3": {"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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 4, "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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "4": {"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": 27, "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": 54, "replicas": 27, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 54, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 54, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 27}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "5": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "6": {"operator": {"Slice": {"axis": 2, "start": 0, "length": 144}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "7": {"operator": {"Slice": {"axis": 2, "start": 144, "length": 144}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "8": {"operator": {"Slice": {"axis": 2, "start": 288, "length": 144}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "9": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "10": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "11": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "12": {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 4, "columns": 1, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "13": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "14": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "15": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "16": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, "17": {"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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 4, "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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "18": {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "19": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "20": {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "21": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "22": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "23": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, "25": {"operator": "LayerNorm", "dispatch": {"LayerNorm": {"parts": 4}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "26": {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "27": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "28": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "29": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "30": {"operator": {"Slice": {"axis": 2, "start": 0, "length": 144}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "31": {"operator": {"Slice": {"axis": 2, "start": 144, "length": 144}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "32": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "33": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "34": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "35": {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 1, "columns": 1, "inner": 1}, "probability_value": null, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "36": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "37": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "38": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "39": {"operator": "LayerNorm", "dispatch": {"LayerNorm": {"parts": 4}}, "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": 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"}}, "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": 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"}}, "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": 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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "40": {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "41": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "42": {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "43": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "44": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "45": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 36}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 36}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 36}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}}, "open_boundaries": [58, 69], "cast_before_copies": [], "packing_rows": null, "parallel_reductions": 0, "disjoint_copy_sources": false, "cast_storage": {"default": "Separate", "operators": {}, "sites": []}}, {"plans": {"0": {"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": 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": 63, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "1": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "2": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "3": {"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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 4, "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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "4": {"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": 27, "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": 54, "replicas": 27, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 54, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 54, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 27}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "5": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "6": {"operator": {"Slice": {"axis": 2, "start": 0, "length": 144}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "7": {"operator": {"Slice": {"axis": 2, "start": 144, "length": 144}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "8": {"operator": {"Slice": {"axis": 2, "start": 288, "length": 144}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "9": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "10": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "11": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "12": {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 4, "columns": 1, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "13": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "14": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "15": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "16": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, "17": {"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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 4, "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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "18": {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "19": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "20": {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "21": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "22": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "23": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, "25": {"operator": "LayerNorm", "dispatch": {"LayerNorm": {"parts": 4}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "26": {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "27": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "28": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "29": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "30": {"operator": {"Slice": {"axis": 2, "start": 0, "length": 144}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "31": {"operator": {"Slice": {"axis": 2, "start": 144, "length": 144}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "32": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "33": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "34": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "35": {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 1, "columns": 1, "inner": 1}, "probability_value": null, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "36": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "37": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "38": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "39": {"operator": "LayerNorm", "dispatch": {"LayerNorm": {"parts": 4}}, "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": 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"}}, "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": 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"}}, "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": 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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "40": {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "41": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "42": {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "43": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "44": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "45": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 36}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 36}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 36}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}}, "open_boundaries": [58, 69], "cast_before_copies": [], "packing_rows": null, "parallel_reductions": 0, "disjoint_copy_sources": false, "cast_storage": {"default": "Separate", "operators": {"3": "ReuseIfSmaller", "4": "ReuseIfSmaller"}, "sites": []}}, {"plans": {"0": {"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": 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": 63, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "1": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "2": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "3": {"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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 4, "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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "4": {"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": 27, "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": 54, "replicas": 27, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 54, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 54, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 27}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "5": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "6": {"operator": {"Slice": {"axis": 2, "start": 0, "length": 144}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "7": {"operator": {"Slice": {"axis": 2, "start": 144, "length": 144}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "8": {"operator": {"Slice": {"axis": 2, "start": 288, "length": 144}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "9": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "10": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "11": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "12": {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 4, "columns": 1, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "13": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "14": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "15": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "16": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, "17": {"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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 4, "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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "18": {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "19": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "21": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "22": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "23": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, "25": {"operator": "LayerNorm", "dispatch": {"LayerNorm": {"parts": 4}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "26": {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "27": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "28": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "29": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "30": {"operator": {"Slice": {"axis": 2, "start": 0, "length": 144}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "31": {"operator": {"Slice": {"axis": 2, "start": 144, "length": 144}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "32": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "33": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "34": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "35": {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 1, "columns": 1, "inner": 1}, "probability_value": null, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "36": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "37": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "38": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "39": {"operator": "LayerNorm", "dispatch": {"LayerNorm": {"parts": 4}}, "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": 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"}}, "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": 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"}}, "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": 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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "40": {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "41": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "42": {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "43": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "44": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "45": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 36}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 36}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 36}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}}, "open_boundaries": [58, 69, 72], "cast_before_copies": [], "packing_rows": null, "parallel_reductions": 0, "disjoint_copy_sources": false, "cast_storage": {"default": "Separate", "operators": {}, "sites": []}}, {"plans": {"0": {"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": 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": 63, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "1": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "2": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "3": {"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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 4, "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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "4": {"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": 27, "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": 54, "replicas": 27, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 54, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 54, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 27}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "5": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "6": {"operator": {"Slice": {"axis": 2, "start": 0, "length": 144}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "7": {"operator": {"Slice": {"axis": 2, "start": 144, "length": 144}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "8": {"operator": {"Slice": {"axis": 2, "start": 288, "length": 144}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "9": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "10": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "11": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "12": {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 4, "columns": 1, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "13": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "14": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "15": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "16": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, "17": {"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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 4, "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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "18": {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "19": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "20": {"operator": "Gelu", "dispatch": {"Pointwise": {"kernel": "Gelu"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "21": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "22": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "23": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, "25": {"operator": "LayerNorm", "dispatch": {"LayerNorm": {"parts": 4}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "26": {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "27": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "28": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "29": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "30": {"operator": {"Slice": {"axis": 2, "start": 0, "length": 144}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "31": {"operator": {"Slice": {"axis": 2, "start": 144, "length": 144}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "32": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "33": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "34": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "35": {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 1, "columns": 1, "inner": 1}, "probability_value": null, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "36": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "37": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "38": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "39": {"operator": "LayerNorm", "dispatch": {"LayerNorm": {"parts": 4}}, "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": 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"}}, "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": 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"}}, "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": 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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "40": {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "41": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "42": {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "43": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "44": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "45": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 36}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 36}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 36}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}}, "open_boundaries": [58, 69, 72], "cast_before_copies": [], "packing_rows": null, "parallel_reductions": 0, "disjoint_copy_sources": false, "cast_storage": {"default": "Separate", "operators": {}, "sites": []}}, {"plans": {"0": {"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": 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": 63, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "1": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "2": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "3": {"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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 4, "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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "4": {"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": 27, "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": 54, "replicas": 27, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 54, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 54, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 27}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "5": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "6": {"operator": {"Slice": {"axis": 2, "start": 0, "length": 144}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "7": {"operator": {"Slice": {"axis": 2, "start": 144, "length": 144}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "8": {"operator": {"Slice": {"axis": 2, "start": 288, "length": 144}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "9": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "10": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "11": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "12": {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 4, "columns": 1, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "13": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "14": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "15": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "16": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, "17": {"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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 4, "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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "18": {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "19": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "21": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "22": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "23": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, "25": {"operator": "LayerNorm", "dispatch": {"LayerNorm": {"parts": 4}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "26": {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "27": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "28": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "29": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "30": {"operator": {"Slice": {"axis": 2, "start": 0, "length": 144}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "31": {"operator": {"Slice": {"axis": 2, "start": 144, "length": 144}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "32": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "33": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "34": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "35": {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 1, "columns": 1, "inner": 1}, "probability_value": null, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "36": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "37": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "38": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "39": {"operator": "LayerNorm", "dispatch": {"LayerNorm": {"parts": 4}}, "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": 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"}}, "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": 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"}}, "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": 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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "40": {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "41": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "42": {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "43": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "44": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "45": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 36}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 36}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 36}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}}, "open_boundaries": [58, 69, 72], "cast_before_copies": [], "packing_rows": null, "parallel_reductions": 0, "disjoint_copy_sources": false, "cast_storage": {"default": "Separate", "operators": {"19": "ReuseIfSmaller", "20": "ReuseIfSmaller"}, "sites": []}}, {"plans": {"0": {"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": 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": 63, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "1": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "2": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "3": {"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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 4, "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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "4": {"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": 27, "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": 54, "replicas": 27, "axes": [{"axis": {"FromEnd": 1}, "partitions": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 96, "column_block": 16}}}, "tiling": {"tile_count": 54, "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": 2, "padding_groups": 1, "block_size": 96, "padding_multiple": 96, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 54, "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": 2, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 27}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "5": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "6": {"operator": {"Slice": {"axis": 2, "start": 0, "length": 144}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "7": {"operator": {"Slice": {"axis": 2, "start": 144, "length": 144}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "8": {"operator": {"Slice": {"axis": 2, "start": 288, "length": 144}}, "dispatch": "View", "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "9": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "10": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "11": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "12": {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 4, "columns": 1, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 1}, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 8, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "13": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "14": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "15": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "16": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, "17": {"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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 4, "replicas": 4, "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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "18": {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "19": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "20": {"operator": "Gelu", "dispatch": {"Pointwise": {"kernel": "Gelu"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "21": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "22": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "23": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}, "25": {"operator": "LayerNorm", "dispatch": {"LayerNorm": {"parts": 4}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 16, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "26": {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "27": {"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": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "replicas": 4, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "28": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "29": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "30": {"operator": {"Slice": {"axis": 2, "start": 0, "length": 144}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "31": {"operator": {"Slice": {"axis": 2, "start": 144, "length": 144}}, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "32": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "33": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "34": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "35": {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [null, null], "query_key": {"rows": 1, "columns": 1, "inner": 1}, "probability_value": null, "materialized": true, "key_block_rows": 64, "padded_query_dimension": 80, "padded_value_dimension": 80}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"Amp": "TransposedRight"}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "column_block": 16}}}, "tiling": {"tile_count": 2, "replicas": 1, "axes": [{"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": 2}, "partitions": 1, "padding_groups": 1, "block_size": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 2}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 1, "block_size": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "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": 2}, {"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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "36": {"operator": {"View": {"split_axis": 2, "merge_axis": 0, "factor": 2, "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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "37": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 32, "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": 32, "padding_multiple": 32, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "38": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "39": {"operator": "LayerNorm", "dispatch": {"LayerNorm": {"parts": 4}}, "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": 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"}}, "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": 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"}}, "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": 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"}}, "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": 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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "40": {"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": 18, "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": 54, "replicas": 18, "axes": [{"axis": {"FromEnd": 1}, "partitions": 3, "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"}}, "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": 18, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 18, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 18}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "41": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "42": {"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"}}, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "43": {"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": 9, "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": 45, "replicas": 9, "axes": [{"axis": {"FromEnd": 1}, "partitions": 5, "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"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": {"F8F143": {"scale_exponent": -4}}, "layout": {"order": {"BlockMajor": {"Matrix": {"row_block": 64, "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": 64, "padding_multiple": 64, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": {"Amp": "Left"}, "tiling": {"tile_count": 9, "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": 1, "padding_groups": 1, "block_size": 2, "padding_multiple": 2, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": 9}]}, "memory_class": "Ipu21Interleaved"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}}, "44": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": "Linear", "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [0]}}}, "45": {"operator": "Add", "dispatch": {"Pointwise": {"kernel": "Add"}}, "requirements": {"inputs": [{"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 36}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 36}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}], "output": {"format": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 36, "replicas": 1, "axes": [{"axis": {"FromEnd": 2}, "partitions": 1, "padding_groups": 1, "block_size": 1, "padding_multiple": 1, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 36}, {"axis": {"FromEnd": 1}, "partitions": 36, "padding_groups": 1, "block_size": 4, "padding_multiple": 4, "shard_padding_multiple": 1, "padding": "Reject", "tile_stride": 1}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": {"MayAliasInputs": [1]}}}}, "open_boundaries": [58, 69, 72], "cast_before_copies": [], "packing_rows": null, "parallel_reductions": 0, "disjoint_copy_sources": false, "cast_storage": {"default": "Separate", "operators": {"19": "ReuseIfSmaller", "20": "ReuseIfSmaller"}, "sites": []}}], "attempts": 4, "mapping": [0, 8, 16, 24, 32, 40, 48, 56, 1, 9, 17, 25, 33, 41, 49, 57, 2, 10, 18, 26, 34, 42, 50, 58, 3, 11, 19, 27, 35, 43, 51, 59, 4, 12, 20, 28, 36, 44, 52, 60, 5, 13, 21, 29, 37, 45, 53, 61, 6, 14, 22, 30, 38, 46, 54, 62, 7, 15, 23, 31, 39, 47, 55, 63], "mapping_checked": true}
