{"version": 1, "context": "ComputeGraph { inputs: [GraphInput { name: \"query\", kind: Host, value: ValueId(0), shape: TensorShape([4, 17, 72]) }, GraphInput { name: \"key\", kind: Host, value: ValueId(1), shape: TensorShape([4, 19, 72]) }, GraphInput { name: \"value\", kind: Host, value: ValueId(2), shape: TensorShape([4, 19, 72]) }], sequences: [], operations: [Operation { id: OperationId(0), inputs: [ValueId(0), ValueId(1), ValueId(2)], results: [ValueId(3)], kind: FlashAttention(AttentionOptions { causal: false, scale: InverseSqrtQueryWidth }) }], outputs: [ValueId(3)], values: {ValueId(0), ValueId(1), ValueId(2), ValueId(3)}, shapes: {ValueId(0): TensorShape([4, 17, 72]), ValueId(1): TensorShape([4, 19, 72]), ValueId(2): TensorShape([4, 19, 72]), ValueId(3): TensorShape([4, 17, 72])}, next_operation: 1, next_value: 4, next_sequence: 0 }\nPipelineConfig { tile_count: 64, optimization_steps: 0, load_search_state: None, save_search_state: None, capacity_baseline: false, exchange_stream_words: None, inputs: {ValueId(0): TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 64, 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: Some(4) }, AxisTiling { axis: FromEnd(3), partitions: 4, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: Some(1) }, AxisTiling { axis: FromEnd(1), partitions: 1, padding_groups: 1, block_size: 16, padding_multiple: 16, shard_padding_multiple: 1, padding: Zero, tile_stride: None }] }, memory_class: Ipu21Standard } }, ValueId(1): TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 4, 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: Some(4) }, AxisTiling { axis: FromEnd(3), partitions: 4, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: Some(1) }, AxisTiling { axis: FromEnd(1), partitions: 1, padding_groups: 1, block_size: 16, padding_multiple: 16, shard_padding_multiple: 1, padding: Zero, tile_stride: None }] }, memory_class: Ipu21Standard } }, ValueId(2): TensorFormat { precision: F16, layout: Layout { order: RowMajor, tiling: TensorTiling { tile_count: 4, 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: Some(4) }, AxisTiling { axis: FromEnd(3), partitions: 4, padding_groups: 1, block_size: 1, padding_multiple: 1, shard_padding_multiple: 1, padding: Reject, tile_stride: Some(1) }, AxisTiling { axis: FromEnd(1), partitions: 1, padding_groups: 1, block_size: 16, padding_multiple: 16, shard_padding_multiple: 1, padding: Zero, tile_stride: None }] }, memory_class: Ipu21Standard } }}, automatic_inputs: {}, operator_candidates: [], 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: Never, attention_strategy: Materialized, attention_products: Independent, attention_fp8_scales: [Some(-4), None] }\nNone", "recipe": {"plans": {"0": {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 6, "columns": 2, "inner": 1}, "probability_value": {"rows": 5, "columns": 3, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}}, "open_boundaries": [], "cast_before_copies": [[0, 0], [0, 1]], "packing_rows": null, "parallel_reductions": 0, "disjoint_copy_sources": false, "in_place_casts": null}, "alternatives": {"0": [{"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 1, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 1, "columns": 2, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 1, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 1, "columns": 2, "inner": 1}, "probability_value": {"rows": 2, "columns": 4, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 1, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 1, "columns": 2, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 1, "columns": 2, "inner": 1}, "probability_value": {"rows": 4, "columns": 4, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 1, "columns": 2, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 1, "columns": 2, "inner": 1}, "probability_value": {"rows": 5, "columns": 3, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 1, "columns": 2, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 1, "columns": 2, "inner": 1}, "probability_value": {"rows": 6, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 1, "columns": 2, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 1, "columns": 2, "inner": 1}, "probability_value": {"rows": 7, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 1, "columns": 2, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 1, "columns": 2, "inner": 1}, "probability_value": {"rows": 8, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 1, "columns": 2, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 1, "columns": 2, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 1, "columns": 2, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 1, "columns": 2, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 1, "columns": 2, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 1, "columns": 2, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 1, "columns": 2, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 1, "columns": 2, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 1, "columns": 2, "inner": 1}, "probability_value": {"rows": 16, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 2, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 2, "columns": 2, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 2, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 2, "columns": 2, "inner": 1}, "probability_value": {"rows": 2, "columns": 4, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 2, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 2, "columns": 2, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 2, "columns": 2, "inner": 1}, "probability_value": {"rows": 4, "columns": 4, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 2, "columns": 2, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 2, "columns": 2, "inner": 1}, "probability_value": {"rows": 5, "columns": 3, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 2, "columns": 2, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 2, "columns": 2, "inner": 1}, "probability_value": {"rows": 6, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 2, "columns": 2, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 2, "columns": 2, "inner": 1}, "probability_value": {"rows": 7, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 2, "columns": 2, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 2, "columns": 2, "inner": 1}, "probability_value": {"rows": 8, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 2, "columns": 2, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 2, "columns": 2, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 2, "columns": 2, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 2, "columns": 2, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 2, "columns": 2, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 2, "columns": 2, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 2, "columns": 2, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 2, "columns": 2, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 2, "columns": 2, "inner": 1}, "probability_value": {"rows": 16, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 3, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 3, "columns": 2, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 3, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 3, "columns": 2, "inner": 1}, "probability_value": {"rows": 2, "columns": 4, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 3, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 3, "columns": 2, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 3, "columns": 2, "inner": 1}, "probability_value": {"rows": 4, "columns": 4, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 3, "columns": 2, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 3, "columns": 2, "inner": 1}, "probability_value": {"rows": 5, "columns": 3, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 3, "columns": 2, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 3, "columns": 2, "inner": 1}, "probability_value": {"rows": 6, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 3, "columns": 2, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 3, "columns": 2, "inner": 1}, "probability_value": {"rows": 7, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 3, "columns": 2, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 3, "columns": 2, "inner": 1}, "probability_value": {"rows": 8, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 3, "columns": 2, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 3, "columns": 2, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 3, "columns": 2, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 3, "columns": 2, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 3, "columns": 2, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 3, "columns": 2, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 3, "columns": 2, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 3, "columns": 2, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 3, "columns": 2, "inner": 1}, "probability_value": {"rows": 16, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 4, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 4, "columns": 2, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 4, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 4, "columns": 2, "inner": 1}, "probability_value": {"rows": 2, "columns": 4, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 4, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 4, "columns": 2, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 4, "columns": 2, "inner": 1}, "probability_value": {"rows": 4, "columns": 4, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 4, "columns": 2, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 4, "columns": 2, "inner": 1}, "probability_value": {"rows": 5, "columns": 3, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 4, "columns": 2, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 4, "columns": 2, "inner": 1}, "probability_value": {"rows": 6, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 4, "columns": 2, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 4, "columns": 2, "inner": 1}, "probability_value": {"rows": 7, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 4, "columns": 2, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 4, "columns": 2, "inner": 1}, "probability_value": {"rows": 8, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 4, "columns": 2, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 4, "columns": 2, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 4, "columns": 2, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 4, "columns": 2, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 4, "columns": 2, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 4, "columns": 2, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 4, "columns": 2, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 4, "columns": 2, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 4, "columns": 2, "inner": 1}, "probability_value": {"rows": 16, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 5, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 5, "columns": 2, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 5, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 5, "columns": 2, "inner": 1}, "probability_value": {"rows": 2, "columns": 4, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 5, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 5, "columns": 2, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 5, "columns": 2, "inner": 1}, "probability_value": {"rows": 4, "columns": 4, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 5, "columns": 2, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 5, "columns": 2, "inner": 1}, "probability_value": {"rows": 5, "columns": 3, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 5, "columns": 2, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 5, "columns": 2, "inner": 1}, "probability_value": {"rows": 6, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 5, "columns": 2, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 5, "columns": 2, "inner": 1}, "probability_value": {"rows": 7, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 5, "columns": 2, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 5, "columns": 2, "inner": 1}, "probability_value": {"rows": 8, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 5, "columns": 2, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 5, "columns": 2, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 5, "columns": 2, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 5, "columns": 2, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 5, "columns": 2, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 5, "columns": 2, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 5, "columns": 2, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 5, "columns": 2, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 5, "columns": 2, "inner": 1}, "probability_value": {"rows": 16, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 6, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 6, "columns": 2, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 6, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 6, "columns": 2, "inner": 1}, "probability_value": {"rows": 2, "columns": 4, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 6, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 6, "columns": 2, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 6, "columns": 2, "inner": 1}, "probability_value": {"rows": 4, "columns": 4, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 6, "columns": 2, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 6, "columns": 2, "inner": 1}, "probability_value": {"rows": 5, "columns": 3, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 6, "columns": 2, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 6, "columns": 2, "inner": 1}, "probability_value": {"rows": 6, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 6, "columns": 2, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 6, "columns": 2, "inner": 1}, "probability_value": {"rows": 7, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 6, "columns": 2, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 6, "columns": 2, "inner": 1}, "probability_value": {"rows": 8, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 6, "columns": 2, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 6, "columns": 2, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 6, "columns": 2, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 6, "columns": 2, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 6, "columns": 2, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 6, "columns": 2, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 6, "columns": 2, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 6, "columns": 2, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 6, "columns": 2, "inner": 1}, "probability_value": {"rows": 16, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 7, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 7, "columns": 2, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 7, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 7, "columns": 2, "inner": 1}, "probability_value": {"rows": 2, "columns": 4, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 7, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 7, "columns": 2, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 7, "columns": 2, "inner": 1}, "probability_value": {"rows": 4, "columns": 4, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 7, "columns": 2, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 7, "columns": 2, "inner": 1}, "probability_value": {"rows": 5, "columns": 3, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 7, "columns": 2, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 7, "columns": 2, "inner": 1}, "probability_value": {"rows": 6, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 7, "columns": 2, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 7, "columns": 2, "inner": 1}, "probability_value": {"rows": 7, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 7, "columns": 2, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 7, "columns": 2, "inner": 1}, "probability_value": {"rows": 8, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 7, "columns": 2, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 7, "columns": 2, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 7, "columns": 2, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 7, "columns": 2, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 7, "columns": 2, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 7, "columns": 2, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 7, "columns": 2, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 7, "columns": 2, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 7, "columns": 2, "inner": 1}, "probability_value": {"rows": 16, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 8, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 8, "columns": 2, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 8, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 8, "columns": 2, "inner": 1}, "probability_value": {"rows": 2, "columns": 4, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 8, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 8, "columns": 2, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 8, "columns": 2, "inner": 1}, "probability_value": {"rows": 4, "columns": 4, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 8, "columns": 2, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 8, "columns": 2, "inner": 1}, "probability_value": {"rows": 5, "columns": 3, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 8, "columns": 2, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 8, "columns": 2, "inner": 1}, "probability_value": {"rows": 6, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 8, "columns": 2, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 8, "columns": 2, "inner": 1}, "probability_value": {"rows": 7, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 8, "columns": 2, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 8, "columns": 2, "inner": 1}, "probability_value": {"rows": 8, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 8, "columns": 2, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 8, "columns": 2, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 8, "columns": 2, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 8, "columns": 2, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 8, "columns": 2, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 8, "columns": 2, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 8, "columns": 2, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 8, "columns": 2, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 8, "columns": 2, "inner": 1}, "probability_value": {"rows": 16, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 9, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 9, "columns": 1, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 9, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 9, "columns": 1, "inner": 1}, "probability_value": {"rows": 2, "columns": 4, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 9, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 9, "columns": 1, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 9, "columns": 1, "inner": 1}, "probability_value": {"rows": 4, "columns": 4, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 9, "columns": 1, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 9, "columns": 1, "inner": 1}, "probability_value": {"rows": 5, "columns": 3, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 9, "columns": 1, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 9, "columns": 1, "inner": 1}, "probability_value": {"rows": 6, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 9, "columns": 1, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 9, "columns": 1, "inner": 1}, "probability_value": {"rows": 7, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 9, "columns": 1, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 9, "columns": 1, "inner": 1}, "probability_value": {"rows": 8, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 9, "columns": 1, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 9, "columns": 1, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 9, "columns": 1, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 9, "columns": 1, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 9, "columns": 1, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 9, "columns": 1, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 9, "columns": 1, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 9, "columns": 1, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 9, "columns": 1, "inner": 1}, "probability_value": {"rows": 16, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 10, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 10, "columns": 1, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 10, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 10, "columns": 1, "inner": 1}, "probability_value": {"rows": 2, "columns": 4, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 10, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 10, "columns": 1, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 10, "columns": 1, "inner": 1}, "probability_value": {"rows": 4, "columns": 4, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 10, "columns": 1, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 10, "columns": 1, "inner": 1}, "probability_value": {"rows": 5, "columns": 3, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 10, "columns": 1, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 10, "columns": 1, "inner": 1}, "probability_value": {"rows": 6, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 10, "columns": 1, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 10, "columns": 1, "inner": 1}, "probability_value": {"rows": 7, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 10, "columns": 1, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 10, "columns": 1, "inner": 1}, "probability_value": {"rows": 8, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 10, "columns": 1, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 10, "columns": 1, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 10, "columns": 1, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 10, "columns": 1, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 10, "columns": 1, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 10, "columns": 1, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 10, "columns": 1, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 10, "columns": 1, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 10, "columns": 1, "inner": 1}, "probability_value": {"rows": 16, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 11, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 11, "columns": 1, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 11, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 11, "columns": 1, "inner": 1}, "probability_value": {"rows": 2, "columns": 4, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 11, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 11, "columns": 1, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 11, "columns": 1, "inner": 1}, "probability_value": {"rows": 4, "columns": 4, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 11, "columns": 1, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 11, "columns": 1, "inner": 1}, "probability_value": {"rows": 5, "columns": 3, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 11, "columns": 1, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 11, "columns": 1, "inner": 1}, "probability_value": {"rows": 6, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 11, "columns": 1, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 11, "columns": 1, "inner": 1}, "probability_value": {"rows": 7, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 11, "columns": 1, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 11, "columns": 1, "inner": 1}, "probability_value": {"rows": 8, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 11, "columns": 1, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 11, "columns": 1, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 11, "columns": 1, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 11, "columns": 1, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 11, "columns": 1, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 11, "columns": 1, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 11, "columns": 1, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 11, "columns": 1, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 11, "columns": 1, "inner": 1}, "probability_value": {"rows": 16, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 12, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 12, "columns": 1, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 12, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 12, "columns": 1, "inner": 1}, "probability_value": {"rows": 2, "columns": 4, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 12, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 12, "columns": 1, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 12, "columns": 1, "inner": 1}, "probability_value": {"rows": 4, "columns": 4, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 12, "columns": 1, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 12, "columns": 1, "inner": 1}, "probability_value": {"rows": 5, "columns": 3, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 12, "columns": 1, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 12, "columns": 1, "inner": 1}, "probability_value": {"rows": 6, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 12, "columns": 1, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 12, "columns": 1, "inner": 1}, "probability_value": {"rows": 7, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 12, "columns": 1, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 12, "columns": 1, "inner": 1}, "probability_value": {"rows": 8, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 12, "columns": 1, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 12, "columns": 1, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 12, "columns": 1, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 12, "columns": 1, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 12, "columns": 1, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 12, "columns": 1, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 12, "columns": 1, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 12, "columns": 1, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 12, "columns": 1, "inner": 1}, "probability_value": {"rows": 16, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 13, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 13, "columns": 1, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 13, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 13, "columns": 1, "inner": 1}, "probability_value": {"rows": 2, "columns": 4, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 13, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 13, "columns": 1, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 13, "columns": 1, "inner": 1}, "probability_value": {"rows": 4, "columns": 4, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 13, "columns": 1, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 13, "columns": 1, "inner": 1}, "probability_value": {"rows": 5, "columns": 3, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 13, "columns": 1, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 13, "columns": 1, "inner": 1}, "probability_value": {"rows": 6, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 13, "columns": 1, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 13, "columns": 1, "inner": 1}, "probability_value": {"rows": 7, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 13, "columns": 1, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 13, "columns": 1, "inner": 1}, "probability_value": {"rows": 8, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 13, "columns": 1, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 13, "columns": 1, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 13, "columns": 1, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 13, "columns": 1, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 13, "columns": 1, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 13, "columns": 1, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 13, "columns": 1, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 13, "columns": 1, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 13, "columns": 1, "inner": 1}, "probability_value": {"rows": 16, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 14, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 14, "columns": 1, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 14, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 14, "columns": 1, "inner": 1}, "probability_value": {"rows": 2, "columns": 4, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 14, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 14, "columns": 1, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 14, "columns": 1, "inner": 1}, "probability_value": {"rows": 4, "columns": 4, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 14, "columns": 1, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 14, "columns": 1, "inner": 1}, "probability_value": {"rows": 5, "columns": 3, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 14, "columns": 1, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 14, "columns": 1, "inner": 1}, "probability_value": {"rows": 6, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 14, "columns": 1, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 14, "columns": 1, "inner": 1}, "probability_value": {"rows": 7, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 14, "columns": 1, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 14, "columns": 1, "inner": 1}, "probability_value": {"rows": 8, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 14, "columns": 1, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 14, "columns": 1, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 14, "columns": 1, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 14, "columns": 1, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 14, "columns": 1, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 14, "columns": 1, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 14, "columns": 1, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 14, "columns": 1, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 14, "columns": 1, "inner": 1}, "probability_value": {"rows": 16, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 15, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 15, "columns": 1, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 15, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 15, "columns": 1, "inner": 1}, "probability_value": {"rows": 2, "columns": 4, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 15, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 15, "columns": 1, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 15, "columns": 1, "inner": 1}, "probability_value": {"rows": 4, "columns": 4, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 15, "columns": 1, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 15, "columns": 1, "inner": 1}, "probability_value": {"rows": 5, "columns": 3, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 15, "columns": 1, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 15, "columns": 1, "inner": 1}, "probability_value": {"rows": 6, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 15, "columns": 1, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 15, "columns": 1, "inner": 1}, "probability_value": {"rows": 7, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 15, "columns": 1, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 15, "columns": 1, "inner": 1}, "probability_value": {"rows": 8, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 15, "columns": 1, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 15, "columns": 1, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 15, "columns": 1, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 15, "columns": 1, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 15, "columns": 1, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 15, "columns": 1, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 15, "columns": 1, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 15, "columns": 1, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 15, "columns": 1, "inner": 1}, "probability_value": {"rows": 16, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 16, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 16, "columns": 1, "inner": 1}, "probability_value": {"rows": 1, "columns": 5, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 16, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 16, "columns": 1, "inner": 1}, "probability_value": {"rows": 2, "columns": 4, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 16, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 16, "columns": 1, "inner": 1}, "probability_value": {"rows": 3, "columns": 2, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 16, "columns": 1, "inner": 1}, "probability_value": {"rows": 4, "columns": 4, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 16, "columns": 1, "inner": 1}, "probability_value": {"rows": 4, "columns": 2, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 16, "columns": 1, "inner": 1}, "probability_value": {"rows": 5, "columns": 3, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 16, "columns": 1, "inner": 1}, "probability_value": {"rows": 5, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 16, "columns": 1, "inner": 1}, "probability_value": {"rows": 6, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 16, "columns": 1, "inner": 1}, "probability_value": {"rows": 6, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 16, "columns": 1, "inner": 1}, "probability_value": {"rows": 7, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 16, "columns": 1, "inner": 1}, "probability_value": {"rows": 7, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 16, "columns": 1, "inner": 1}, "probability_value": {"rows": 8, "columns": 2, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 16, "columns": 1, "inner": 1}, "probability_value": {"rows": 8, "columns": 1, "inner": 2}, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 16, "columns": 1, "inner": 1}, "probability_value": {"rows": 9, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 16, "columns": 1, "inner": 1}, "probability_value": {"rows": 10, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 16, "columns": 1, "inner": 1}, "probability_value": {"rows": 11, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 16, "columns": 1, "inner": 1}, "probability_value": {"rows": 12, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 16, "columns": 1, "inner": 1}, "probability_value": {"rows": 13, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 16, "columns": 1, "inner": 1}, "probability_value": {"rows": 14, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 16, "columns": 1, "inner": 1}, "probability_value": {"rows": 15, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}, {"operator": {"FlashAttention": {"options": {"causal": false, "scale": "InverseSqrtQueryWidth"}, "accumulate": "F32"}}, "dispatch": {"Attention": {"fp8_scales": [-4, null], "query_key": {"rows": 16, "columns": 1, "inner": 1}, "probability_value": {"rows": 16, "columns": 1, "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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 4, "replicas": 1, "axes": [{"axis": {"FromEnd": 3}, "partitions": 4, "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": 4}, {"axis": {"FromEnd": 1}, "partitions": 1, "padding_groups": 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": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "local_staging": "Direct", "materialization": "Complete"}, "output_aliasing": "Fresh"}, "deferred_output": null}]}, "inputs": {"0": {"precision": "F16", "layout": {"order": "RowMajor", "tiling": {"tile_count": 64, "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": 4}, {"axis": {"FromEnd": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}, "1": {"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": 3}, "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": 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": 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": 3}, "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": 16, "padding_multiple": 16, "shard_padding_multiple": 1, "padding": "Zero", "tile_stride": null}]}, "memory_class": "Ipu21Standard"}}}, "visited": [], "mapping": null, "mapping_checked": false, "attempts": 0}