{ "fixtureArrays": { "ort_batch2_bias_flattened_tokens_input_skipT": [0.1, -0.2, 0.3, 1, 0.5, 0.1, 0.4, 1.6, 1.8, -0.3, 0, 1, -0.5, 0.4, 0.8, -0.6], "ort_batch2_flattened_tokens_input_inputT": [0.8, -0.5, 0, 1, 0.5, 0.2, 0.3, -0.6, 0.8, -0.5, 0, 1, 0.5, 0.2, 0.3, -0.6], "ort_batch2_skip_broadcast_no_batch_size_output_outputT": [0.2843385934829712, -0.17090578377246857, -0.9289716482162476, 4.692415237426758, 0.4611165225505829, -0.21333980560302734, -0.29631003737449646, 3.5148544311523438, 0.2843385934829712, -0.17090578377246857, -0.9289716482162476, 4.692415237426758, 0.4611165225505829, -0.21333980560302734, -0.29631003737449646, 3.5148544311523438] }, "cases": [ { "name": "rank3_exact_skip_shape", "provenance": { "source": "onnxruntime/contrib_ops/cpu/bert/skip_layer_norm.cc", "notes": "Pins the public rank-3 same-shape skip mode independently of the two documented broadcast forms." }, "attrs": { "epsilon": 0.00001 }, "inputs": { "inputT": { "dtype": "float32", "shape": [2, 2, 4], "data": { "kind": "linspace", "start": -1.5, "end": 1.5 } }, "skipT": { "dtype": "float32", "shape": [2, 2, 4], "data": { "kind": "linspace", "start": 0.75, "end": -0.25 } }, "gammaT": { "dtype": "float32", "shape": [4], "data": { "kind": "values", "values": [1.0, 0.5, 1.5, -0.25] } }, "betaT": { "dtype": "float32", "shape": [4], "data": { "kind": "values", "values": [0.1, -0.2, 0.3, -0.4] } } }, "outputs": { "outputT": { "dtype": "float32", "shape": [2, 2, 4], "tolerance": 0.00002 } } }, { "name": "no_bias", "attrs": { "epsilon": 0.00001 }, "inputs": { "inputT": { "dtype": "float32", "shape": [3, 8], "data": { "kind": "fillFloat32", "sinStep": 0.17, "cosStep": 0.31 } }, "skipT": { "dtype": "float32", "shape": [3, 8], "data": { "kind": "fillFloat32", "sinStep": 0.11, "cosStep": 0.23 } }, "gammaT": { "dtype": "float32", "shape": [8], "data": { "kind": "fillFloat32", "sinStep": 0.07, "cosStep": 0.41, "scale": 0.2 } }, "betaT": { "dtype": "float32", "shape": [8], "data": { "kind": "fillFloat32", "sinStep": 0.19, "cosStep": 0.13, "scale": 0.1 } } }, "outputs": { "outputT": { "dtype": "float32", "shape": [3, 8], "tolerance": 0.00002 }, "residualT": { "dtype": "float32", "shape": [3, 8], "tolerance": 0.000001 } } }, { "name": "f32_epsilon_zero_explicit_tiny_variance", "attrs": { "epsilon": 0 }, "provenance": { "source": "onnxruntime/contrib_ops/webgpu/bert/skip_layer_norm.h", "test": "GetAttrOrDefault epsilon semantics", "notes": "An explicit epsilon=0.0 must be honored, not replaced by the 1e-12 schema default via a truthiness fallback. The 1e-7-scale rows make that difference numerically observable." }, "inputs": { "inputT": { "dtype": "float32", "shape": [3, 8], "data": { "kind": "fillFloat32", "sinStep": 0.17, "cosStep": 0.31, "scale": 1e-7 } }, "skipT": { "dtype": "float32", "shape": [3, 8], "data": { "kind": "fillFloat32", "sinStep": 0.11, "cosStep": 0.23, "scale": 1e-7 } }, "gammaT": { "dtype": "float32", "shape": [8], "data": { "kind": "fillFloat32", "sinStep": 0.07, "cosStep": 0.41, "scale": 0.2 } }, "betaT": { "dtype": "float32", "shape": [8], "data": { "kind": "fillFloat32", "sinStep": 0.19, "cosStep": 0.13, "scale": 0.1 } } }, "outputs": { "outputT": { "dtype": "float32", "shape": [3, 8], "tolerance": 0.0001 }, "residualT": { "dtype": "float32", "shape": [3, 8], "tolerance": 1e-9 } } }, { "name": "bias", "attrs": { "epsilon": 0.00001 }, "inputs": { "inputT": { "dtype": "float32", "shape": [3, 8], "data": { "kind": "fillFloat32", "sinStep": 0.17, "cosStep": 0.31 } }, "skipT": { "dtype": "float32", "shape": [3, 8], "data": { "kind": "fillFloat32", "sinStep": 0.11, "cosStep": 0.23 } }, "gammaT": { "dtype": "float32", "shape": [8], "data": { "kind": "fillFloat32", "sinStep": 0.07, "cosStep": 0.41, "scale": 0.2 } }, "betaT": { "dtype": "float32", "shape": [8], "data": { "kind": "fillFloat32", "sinStep": 0.19, "cosStep": 0.13, "scale": 0.1 } }, "biasT": { "dtype": "float32", "shape": [8], "data": { "kind": "fillFloat32", "sinStep": 0.29, "cosStep": 0.17, "scale": 0.08 } } }, "outputs": { "outputT": { "dtype": "float32", "shape": [3, 8], "tolerance": 0.00002 }, "residualT": { "dtype": "float32", "shape": [3, 8], "tolerance": 0.000001 } } }, { "name": "zero_variance_returns_beta", "attrs": { "epsilon": 0.00001 }, "inputs": { "inputT": { "dtype": "float32", "shape": [2, 4], "data": { "kind": "values", "values": [5.0, 5.0, 5.0, 5.0, -3.0, -3.0, -3.0, -3.0] } }, "skipT": { "dtype": "float32", "shape": [2, 4], "data": { "kind": "values", "values": [1.0, 1.0, 1.0, 1.0, 7.0, 7.0, 7.0, 7.0] } }, "gammaT": { "dtype": "float32", "shape": [4], "data": { "kind": "values", "values": [10.0, -2.0, 3.0, 4.0] } }, "betaT": { "dtype": "float32", "shape": [4], "data": { "kind": "values", "values": [0.5, -1.0, 2.0, -3.0] } } }, "outputs": { "outputT": { "dtype": "float32", "shape": [2, 4], "tolerance": 0.000001 }, "residualT": { "dtype": "float32", "shape": [2, 4], "tolerance": 0.000001 } } }, { "name": "hidden_size_one_bias_path", "attrs": { "epsilon": 0.00001 }, "inputs": { "inputT": { "dtype": "float32", "shape": [3, 1], "data": { "kind": "values", "values": [2.0, -4.0, 0.5] } }, "skipT": { "dtype": "float32", "shape": [3, 1], "data": { "kind": "values", "values": [3.0, 1.0, -0.5] } }, "gammaT": { "dtype": "float32", "shape": [1], "data": { "kind": "values", "values": [7.0] } }, "betaT": { "dtype": "float32", "shape": [1], "data": { "kind": "values", "values": [-2.0] } }, "biasT": { "dtype": "float32", "shape": [1], "data": { "kind": "values", "values": [0.25] } } }, "outputs": { "outputT": { "dtype": "float32", "shape": [3, 1], "tolerance": 0.000001 }, "residualT": { "dtype": "float32", "shape": [3, 1], "tolerance": 0.000001 } } }, { "name": "hidden_size_one_bias_output_only", "provenance": { "notes": "Pins the scalar hidden-size-one closed form when the optional residual sum is not requested." }, "attrs": { "epsilon": 0.00001 }, "inputs": { "inputT": { "dtype": "float32", "shape": [1, 1], "data": { "kind": "values", "values": [2.0] } }, "skipT": { "dtype": "float32", "shape": [1, 1], "data": { "kind": "values", "values": [3.0] } }, "gammaT": { "dtype": "float32", "shape": [1], "data": { "kind": "values", "values": [7.0] } }, "betaT": { "dtype": "float32", "shape": [1], "data": { "kind": "values", "values": [-2.0] } }, "biasT": { "dtype": "float32", "shape": [1], "data": { "kind": "values", "values": [0.25] } } }, "outputs": { "outputT": { "dtype": "float32", "shape": [1, 1], "tolerance": 0.000001 } } }, { "name": "hidden_size_one_no_beta_output_only", "provenance": { "notes": "Pins the supported no-beta hidden-size-one closed form: every centered value is zero, so the output is zero." }, "attrs": { "epsilon": 0.00001 }, "inputs": { "inputT": { "dtype": "float32", "shape": [1, 1], "data": { "kind": "values", "values": [2.0] } }, "skipT": { "dtype": "float32", "shape": [1, 1], "data": { "kind": "values", "values": [3.0] } }, "gammaT": { "dtype": "float32", "shape": [1], "data": { "kind": "values", "values": [7.0] } } }, "outputs": { "outputT": { "dtype": "float32", "shape": [1, 1], "tolerance": 0.000001 } } }, { "name": "hidden_size_one_bias_rows65535_dispatch_edge", "provenance": { "notes": "Hidden size 1 across 65,535 rows exercises the maximum single-dimension workgroup count with one value per normalization row." }, "attrs": { "epsilon": 0.00001 }, "inputs": { "inputT": { "dtype": "float32", "shape": [65535, 1], "data": { "kind": "fillFloat32", "scale": 0.2, "sinStep": 0.017, "cosStep": 0.031 } }, "skipT": { "dtype": "float32", "shape": [65535, 1], "data": { "kind": "fillFloat32", "scale": 0.2, "sinStep": 0.011, "cosStep": 0.023 } }, "gammaT": { "dtype": "float32", "shape": [1], "data": { "kind": "values", "values": [1.125] } }, "betaT": { "dtype": "float32", "shape": [1], "data": { "kind": "values", "values": [-0.25] } }, "biasT": { "dtype": "float32", "shape": [1], "data": { "kind": "values", "values": [0.125] } } }, "outputs": { "outputT": { "dtype": "float32", "shape": [65535, 1], "tolerance": 0.000001 }, "residualT": { "dtype": "float32", "shape": [65535, 1], "tolerance": 0.000001 } } }, { "name": "large_mean_small_variance_centered", "attrs": { "epsilon": 0.00001 }, "inputs": { "inputT": { "dtype": "float32", "shape": [1, 4], "data": { "kind": "values", "values": [40000.0, 40001.0, 40002.0, 40003.0] } }, "skipT": { "dtype": "float32", "shape": [1, 4], "data": { "kind": "values", "values": [-39999.0, -39999.5, -40000.0, -40000.5] } }, "gammaT": { "dtype": "float32", "shape": [4], "data": { "kind": "values", "values": [1.0, 2.0, -1.0, 0.5] } }, "betaT": { "dtype": "float32", "shape": [4], "data": { "kind": "values", "values": [0.0, 0.5, -0.25, 1.0] } } }, "outputs": { "outputT": { "dtype": "float32", "shape": [1, 4], "tolerance": 0.000001 }, "residualT": { "dtype": "float32", "shape": [1, 4], "tolerance": 0.000001 } } }, { "name": "ort_zero_tokens_null_input", "provenance": { "source": "onnxruntime/test/contrib_ops/skiplayernorm_op_test.cc", "test": "SkipLayerNormTest.SkipLayerNormNullInput", "notes": "ORT shape [1, 0, 4] is represented as lowered token rows [0, 4]." }, "attrs": { "epsilon": 1e-12 }, "inputs": { "inputT": { "dtype": "float32", "shape": [0, 4], "data": { "kind": "values", "values": [] } }, "skipT": { "dtype": "float32", "shape": [0, 4], "data": { "kind": "values", "values": [] } }, "gammaT": { "dtype": "float32", "shape": [4], "data": { "kind": "values", "values": [0.3, 0.2, 4.0, 2.2] } }, "betaT": { "dtype": "float32", "shape": [4], "data": { "kind": "values", "values": [0.2, 0.1, 0.4, 1.6] } } }, "outputs": { "outputT": { "dtype": "float32", "shape": [0, 4], "tolerance": 0 }, "residualT": { "dtype": "float32", "shape": [0, 4], "tolerance": 0 } } }, { "name": "ort_batch1_flattened_tokens", "provenance": { "source": "onnxruntime/test/contrib_ops/skiplayernorm_op_test.cc", "test": "SkipLayerNormTest.SkipLayerNormBatch1", "notes": "This package flattens ORT shape [1, 2, 4] to two four-wide token rows. Epsilon is omitted to exercise the schema default of 1e-12." }, "inputs": { "inputT": { "dtype": "float32", "shape": [2, 4], "data": { "kind": "values", "values": [0.8, -0.5, 0.0, 1.0, 0.5, 0.2, 0.3, -0.6] } }, "skipT": { "dtype": "float32", "shape": [2, 4], "data": { "kind": "values", "values": [0.1, -0.2, 0.3, 1.0, 0.5, 0.1, 0.4, 1.6] } }, "gammaT": { "dtype": "float32", "shape": [4], "data": { "kind": "values", "values": [0.3, 0.2, 4.0, 2.2] } }, "betaT": { "dtype": "float32", "shape": [4], "data": { "kind": "values", "values": [0.2, 0.1, 0.4, 1.6] } } }, "outputs": { "outputT": { "dtype": "float32", "shape": [2, 4], "tolerance": 0.00002 }, "residualT": { "dtype": "float32", "shape": [2, 4], "tolerance": 0.000001 } } }, { "name": "ort_batch2_bias_flattened_tokens", "provenance": { "source": "onnxruntime/test/contrib_ops/skiplayernorm_op_test.cc", "test": "SkipLayerNormTest.SkipLayerNormBatch2_Bias", "notes": "This package flattens ORT shape [2, 2, 4] to four four-wide token rows." }, "attrs": { "epsilon": 1e-12 }, "inputs": { "inputT": { "dtype": "float32", "shape": [4, 4], "data": { "kind": "values", "values": [0.7, -0.4, -0.2, 1.2, 0.4, 0.3, 0.1, -0.4, 0.7, -0.4, -0.2, 1.2, 0.4, 0.3, 0.1, -0.4] } }, "skipT": { "dtype": "float32", "shape": [4, 4], "data": { "kind": "values", "values": { "$ref": "#/fixtureArrays/ort_batch2_bias_flattened_tokens_input_skipT" } } }, "gammaT": { "dtype": "float32", "shape": [4], "data": { "kind": "values", "values": [0.3, 0.2, 4.0, 2.2] } }, "betaT": { "dtype": "float32", "shape": [4], "data": { "kind": "values", "values": [0.2, 0.1, 0.4, 1.6] } }, "biasT": { "dtype": "float32", "shape": [4], "data": { "kind": "values", "values": [0.1, -0.1, 0.2, -0.2] } } }, "outputs": { "outputT": { "dtype": "float32", "shape": [4, 4], "tolerance": 0.00002 }, "residualT": { "dtype": "float32", "shape": [4, 4], "tolerance": 0.000001 } } }, { "name": "ort_batch2_flattened_tokens", "provenance": { "source": "onnxruntime/test/contrib_ops/skiplayernorm_op_test.cc", "test": "SkipLayerNormTest.SkipLayerNormBatch2", "notes": "This package flattens ORT shape [2, 2, 4] to four four-wide token rows." }, "attrs": { "epsilon": 1e-12 }, "inputs": { "inputT": { "dtype": "float32", "shape": [4, 4], "data": { "kind": "values", "values": { "$ref": "#/fixtureArrays/ort_batch2_flattened_tokens_input_inputT" } } }, "skipT": { "dtype": "float32", "shape": [4, 4], "data": { "kind": "values", "values": { "$ref": "#/fixtureArrays/ort_batch2_bias_flattened_tokens_input_skipT" } } }, "gammaT": { "dtype": "float32", "shape": [4], "data": { "kind": "values", "values": [0.3, 0.2, 4.0, 2.2] } }, "betaT": { "dtype": "float32", "shape": [4], "data": { "kind": "values", "values": [0.2, 0.1, 0.4, 1.6] } } }, "outputs": { "outputT": { "dtype": "float32", "shape": [4, 4], "tolerance": 0.00002 }, "residualT": { "dtype": "float32", "shape": [4, 4], "tolerance": 0.000001 } } }, { "name": "large_hidden_320_no_bias", "attrs": { "epsilon": 0.00001 }, "inputs": { "inputT": { "dtype": "float32", "shape": [2, 320], "data": { "kind": "fillFloat32", "scale": 0.2, "sinStep": 0.017, "cosStep": 0.031 } }, "skipT": { "dtype": "float32", "shape": [2, 320], "data": { "kind": "fillFloat32", "scale": 0.2, "sinStep": 0.011, "cosStep": 0.023 } }, "gammaT": { "dtype": "float32", "shape": [320], "data": { "kind": "fillFloat32", "scale": 0.1, "offset": 1.0, "sinStep": 0.007, "cosStep": 0.041 } }, "betaT": { "dtype": "float32", "shape": [320], "data": { "kind": "fillFloat32", "scale": 0.1, "sinStep": 0.019, "cosStep": 0.013 } } }, "outputs": { "outputT": { "dtype": "float32", "shape": [2, 320], "tolerance": 0.0002 }, "residualT": { "dtype": "float32", "shape": [2, 320], "tolerance": 0.000001 } } }, { "name": "large_hidden_320_bias", "attrs": { "epsilon": 0.00001 }, "inputs": { "inputT": { "dtype": "float32", "shape": [2, 320], "data": { "kind": "fillFloat32", "scale": 0.2, "sinStep": 0.013, "cosStep": 0.029 } }, "skipT": { "dtype": "float32", "shape": [2, 320], "data": { "kind": "fillFloat32", "scale": 0.2, "sinStep": 0.019, "cosStep": 0.037 } }, "gammaT": { "dtype": "float32", "shape": [320], "data": { "kind": "fillFloat32", "scale": 0.1, "offset": 1.0, "sinStep": 0.029, "cosStep": 0.017 } }, "betaT": { "dtype": "float32", "shape": [320], "data": { "kind": "fillFloat32", "scale": 0.1, "sinStep": 0.023, "cosStep": 0.011 } }, "biasT": { "dtype": "float32", "shape": [320], "data": { "kind": "fillFloat32", "scale": 0.05, "sinStep": 0.031, "cosStep": 0.007 } } }, "outputs": { "outputT": { "dtype": "float32", "shape": [2, 320], "tolerance": 0.0002 }, "residualT": { "dtype": "float32", "shape": [2, 320], "tolerance": 0.000001 } } }, { "name": "ort_batch2_skip_broadcast_no_batch_size", "provenance": { "source": "onnxruntime/test/contrib_ops/skiplayernorm_op_test.cc", "test": "SkipLayerNormTest.SkipLayerNormBatch2_Skip_Broadcast_No_Batch_Size" }, "attrs": { "epsilon": 1e-12 }, "inputs": { "inputT": { "dtype": "float32", "shape": [2, 2, 4], "data": { "kind": "values", "values": { "$ref": "#/fixtureArrays/ort_batch2_flattened_tokens_input_inputT" } } }, "skipT": { "dtype": "float32", "shape": [2, 4], "data": { "kind": "values", "values": [0.1, -0.2, 0.3, 1.0, 0.5, 0.1, 0.4, 1.6] } }, "gammaT": { "dtype": "float32", "shape": [4], "data": { "kind": "values", "values": [0.3, 0.2, 4.0, 2.2] } }, "betaT": { "dtype": "float32", "shape": [4], "data": { "kind": "values", "values": [0.2, 0.1, 0.4, 1.6] } } }, "outputs": { "outputT": { "dtype": "float32", "shape": [2, 2, 4], "tolerance": 0.00002, "data": { "kind": "values", "values": { "$ref": "#/fixtureArrays/ort_batch2_skip_broadcast_no_batch_size_output_outputT" } } } } }, { "name": "ort_batch1_no_beta_flattened_tokens", "provenance": { "source": "onnxruntime/test/contrib_ops/skiplayernorm_op_test.cc", "test": "SkipLayerNormTest.SkipLayerNormBatch1_NoBeta", "notes": "ORT shape [1, 2, 4] is represented as [2, 4] tokens by this lowered fixture, with beta omitted." }, "attrs": { "epsilon": 1e-12 }, "inputs": { "inputT": { "dtype": "float32", "shape": [2, 4], "data": { "kind": "values", "values": [0.8, -0.5, 0.0, 1.0, 0.5, 0.2, 0.3, -0.6] } }, "skipT": { "dtype": "float32", "shape": [2, 4], "data": { "kind": "values", "values": [0.1, -0.2, 0.3, 1.0, 0.5, 0.1, 0.4, 1.6] } }, "gammaT": { "dtype": "float32", "shape": [4], "data": { "kind": "values", "values": [0.3, 0.2, 4.0, 2.2] } } }, "outputs": { "outputT": { "dtype": "float32", "shape": [2, 4], "tolerance": 0.00002, "data": { "kind": "values", "values": [0.0843385934829712, -0.27090578377246854, -1.3289716482162477, 3.092415237426758, 0.2611165225505829, -0.3133398056030273, -0.6963100373744965, 1.9148544311523439] } } } }, { "name": "no_beta_output_only_hidden6_unaligned_row", "provenance": { "notes": "Beta and optional outputs are omitted, and hidden size 6 is not divisible by four, selecting the scalar no-beta output-only row path on every tier." }, "attrs": { "epsilon": 0.00001 }, "inputs": { "inputT": { "dtype": "float32", "shape": [2, 6], "data": { "kind": "fillFloat32", "sinStep": 0.17, "cosStep": 0.31 } }, "skipT": { "dtype": "float32", "shape": [2, 6], "data": { "kind": "fillFloat32", "sinStep": 0.11, "cosStep": 0.23 } }, "gammaT": { "dtype": "float32", "shape": [6], "data": { "kind": "fillFloat32", "sinStep": 0.07, "cosStep": 0.41, "scale": 0.2 } } }, "outputs": { "outputT": { "dtype": "float32", "shape": [2, 6], "tolerance": 0.00002 } } }, { "name": "beta_no_bias_output_only", "provenance": { "notes": "Exercises beta with all three optional auxiliary outputs absent, using a vec4-aligned hidden size of 8." }, "attrs": { "epsilon": 0.00001 }, "inputs": { "inputT": { "dtype": "float32", "shape": [3, 8], "data": { "kind": "fillFloat32", "sinStep": 0.17, "cosStep": 0.31 } }, "skipT": { "dtype": "float32", "shape": [3, 8], "data": { "kind": "fillFloat32", "sinStep": 0.11, "cosStep": 0.23 } }, "gammaT": { "dtype": "float32", "shape": [8], "data": { "kind": "fillFloat32", "sinStep": 0.07, "cosStep": 0.41, "scale": 0.2 } }, "betaT": { "dtype": "float32", "shape": [8], "data": { "kind": "fillFloat32", "sinStep": 0.13, "cosStep": 0.29, "scale": 0.1 } } }, "outputs": { "outputT": { "dtype": "float32", "shape": [3, 8], "tolerance": 0.00002 } } }, { "name": "beta_no_bias_output_only_hidden6_unaligned_row", "provenance": { "notes": "Beta is present, bias and optional outputs are omitted, and hidden size 6 selects the scalar output-only row path on every tier." }, "attrs": { "epsilon": 0.00001 }, "inputs": { "inputT": { "dtype": "float32", "shape": [2, 6], "data": { "kind": "fillFloat32", "sinStep": 0.19, "cosStep": 0.37 } }, "skipT": { "dtype": "float32", "shape": [2, 6], "data": { "kind": "fillFloat32", "sinStep": 0.23, "cosStep": 0.43 } }, "gammaT": { "dtype": "float32", "shape": [6], "data": { "kind": "fillFloat32", "sinStep": 0.07, "cosStep": 0.41, "scale": 0.2 } }, "betaT": { "dtype": "float32", "shape": [6], "data": { "kind": "fillFloat32", "sinStep": 0.13, "cosStep": 0.29, "scale": 0.1 } } }, "outputs": { "outputT": { "dtype": "float32", "shape": [2, 6], "tolerance": 0.00002 } } }, { "name": "beta_bias_output_only", "provenance": { "notes": "bias + beta with no optional outputs uses six storage buffers instead of the nine required when all optional outputs are present. It therefore remains valid at WebGPU's guaranteed minimum of eight storage buffers." }, "attrs": { "epsilon": 0.00001 }, "inputs": { "inputT": { "dtype": "float32", "shape": [3, 8], "data": { "kind": "fillFloat32", "sinStep": 0.17, "cosStep": 0.31 } }, "skipT": { "dtype": "float32", "shape": [3, 8], "data": { "kind": "fillFloat32", "sinStep": 0.11, "cosStep": 0.23 } }, "gammaT": { "dtype": "float32", "shape": [8], "data": { "kind": "fillFloat32", "sinStep": 0.07, "cosStep": 0.41, "scale": 0.2 } }, "betaT": { "dtype": "float32", "shape": [8], "data": { "kind": "fillFloat32", "sinStep": 0.13, "cosStep": 0.29, "scale": 0.1 } }, "biasT": { "dtype": "float32", "shape": [8], "data": { "kind": "fillFloat32", "sinStep": 0.05, "cosStep": 0.47, "scale": 0.3 } } }, "outputs": { "outputT": { "dtype": "float32", "shape": [3, 8], "tolerance": 0.00002 } } }, { "name": "beta_bias_output_only_hidden6_unaligned_row", "provenance": { "notes": "Row-kernel arm of beta_bias_output_only: hidden=6 fails vec4Aligned. 6 storage buffers." }, "attrs": { "epsilon": 0.00001 }, "inputs": { "inputT": { "dtype": "float32", "shape": [2, 6], "data": { "kind": "fillFloat32", "sinStep": 0.19, "cosStep": 0.37 } }, "skipT": { "dtype": "float32", "shape": [2, 6], "data": { "kind": "fillFloat32", "sinStep": 0.23, "cosStep": 0.43 } }, "gammaT": { "dtype": "float32", "shape": [6], "data": { "kind": "fillFloat32", "sinStep": 0.07, "cosStep": 0.41, "scale": 0.2 } }, "betaT": { "dtype": "float32", "shape": [6], "data": { "kind": "fillFloat32", "sinStep": 0.13, "cosStep": 0.29, "scale": 0.1 } }, "biasT": { "dtype": "float32", "shape": [6], "data": { "kind": "fillFloat32", "sinStep": 0.05, "cosStep": 0.47, "scale": 0.3 } } }, "outputs": { "outputT": { "dtype": "float32", "shape": [2, 6], "tolerance": 0.00002 } } }, { "name": "ort_batch2_skip_broadcast_batch_size_one", "provenance": { "source": "onnxruntime/test/contrib_ops/skiplayernorm_op_test.cc", "test": "SkipLayerNormTest.SkipLayerNormBatch2_Skip_Broadcast_Batch_Size_1" }, "attrs": { "epsilon": 1e-12 }, "inputs": { "inputT": { "dtype": "float32", "shape": [2, 2, 4], "data": { "kind": "values", "values": { "$ref": "#/fixtureArrays/ort_batch2_flattened_tokens_input_inputT" } } }, "skipT": { "dtype": "float32", "shape": [1, 2, 4], "data": { "kind": "values", "values": [0.1, -0.2, 0.3, 1.0, 0.5, 0.1, 0.4, 1.6] } }, "gammaT": { "dtype": "float32", "shape": [4], "data": { "kind": "values", "values": [0.3, 0.2, 4.0, 2.2] } }, "betaT": { "dtype": "float32", "shape": [4], "data": { "kind": "values", "values": [0.2, 0.1, 0.4, 1.6] } } }, "outputs": { "outputT": { "dtype": "float32", "shape": [2, 2, 4], "tolerance": 0.00002, "data": { "kind": "values", "values": { "$ref": "#/fixtureArrays/ort_batch2_skip_broadcast_no_batch_size_output_outputT" } } } } }, { "name": "f16_hidden768_bias_residual", "requires": { "features": ["shader-f16"] }, "attrs": { "epsilon": 0.00001 }, "inputs": { "inputT": { "dtype": "float16", "shape": [4, 768], "data": { "kind": "fillFloat32", "scale": 0.2, "sinStep": 0.017, "cosStep": 0.031 } }, "skipT": { "dtype": "float16", "shape": [4, 768], "data": { "kind": "fillFloat32", "scale": 0.2, "sinStep": 0.011, "cosStep": 0.023 } }, "gammaT": { "dtype": "float16", "shape": [768], "data": { "kind": "fillFloat32", "scale": 0.1, "offset": 1.0, "sinStep": 0.007, "cosStep": 0.041 } }, "betaT": { "dtype": "float16", "shape": [768], "data": { "kind": "fillFloat32", "scale": 0.1, "sinStep": 0.019, "cosStep": 0.013 } }, "biasT": { "dtype": "float16", "shape": [768], "data": { "kind": "fillFloat32", "scale": 0.05, "sinStep": 0.031, "cosStep": 0.007 } } }, "outputs": { "outputT": { "dtype": "float16", "shape": [4, 768], "tolerance": 0.01 }, "residualT": { "dtype": "float16", "shape": [4, 768], "tolerance": 0.005 } } }, { "name": "f32_hidden768_no_bias_residual", "provenance": { "notes": "A compact hidden-size-768 residual normalization exercises the subgroup route and its reduced-tier fallback without bias." }, "attrs": { "epsilon": 0.00001 }, "inputs": { "inputT": { "dtype": "float32", "shape": [4, 768], "data": { "kind": "fillFloat32", "scale": 0.2, "sinStep": 0.017, "cosStep": 0.031 } }, "skipT": { "dtype": "float32", "shape": [4, 768], "data": { "kind": "fillFloat32", "scale": 0.2, "sinStep": 0.011, "cosStep": 0.023 } }, "gammaT": { "dtype": "float32", "shape": [768], "data": { "kind": "fillFloat32", "scale": 0.1, "offset": 1.0, "sinStep": 0.007, "cosStep": 0.041 } }, "betaT": { "dtype": "float32", "shape": [768], "data": { "kind": "fillFloat32", "scale": 0.1, "sinStep": 0.019, "cosStep": 0.013 } } }, "outputs": { "outputT": { "dtype": "float32", "shape": [4, 768], "tolerance": 0.0002 }, "residualT": { "dtype": "float32", "shape": [4, 768], "tolerance": 0.000002 } } }, { "name": "f32_hidden2048_bias_residual", "attrs": { "epsilon": 0.00001 }, "inputs": { "inputT": { "dtype": "float32", "shape": [2, 2048], "data": { "kind": "fillFloat32", "scale": 0.2, "sinStep": 0.013, "cosStep": 0.029 } }, "skipT": { "dtype": "float32", "shape": [2, 2048], "data": { "kind": "fillFloat32", "scale": 0.2, "sinStep": 0.019, "cosStep": 0.037 } }, "gammaT": { "dtype": "float32", "shape": [2048], "data": { "kind": "fillFloat32", "scale": 0.1, "offset": 1.0, "sinStep": 0.029, "cosStep": 0.017 } }, "betaT": { "dtype": "float32", "shape": [2048], "data": { "kind": "fillFloat32", "scale": 0.1, "sinStep": 0.023, "cosStep": 0.011 } }, "biasT": { "dtype": "float32", "shape": [2048], "data": { "kind": "fillFloat32", "scale": 0.05, "sinStep": 0.031, "cosStep": 0.007 } } }, "outputs": { "outputT": { "dtype": "float32", "shape": [2, 2048], "tolerance": 0.0005 }, "residualT": { "dtype": "float32", "shape": [2, 2048], "tolerance": 0.000002 } } }, { "name": "f32_hidden1025_bias_residual", "provenance": { "notes": "A compact hidden-size-1025 normalization exercises the unaligned two-pass path with bias and residual output." }, "attrs": { "epsilon": 0.00001 }, "inputs": { "inputT": { "dtype": "float32", "shape": [3, 1025], "data": { "kind": "fillFloat32", "scale": 0.2, "sinStep": 0.013, "cosStep": 0.029 } }, "skipT": { "dtype": "float32", "shape": [3, 1025], "data": { "kind": "fillFloat32", "scale": 0.2, "sinStep": 0.019, "cosStep": 0.037 } }, "gammaT": { "dtype": "float32", "shape": [1025], "data": { "kind": "fillFloat32", "scale": 0.1, "offset": 1.0, "sinStep": 0.029, "cosStep": 0.017 } }, "betaT": { "dtype": "float32", "shape": [1025], "data": { "kind": "fillFloat32", "scale": 0.1, "sinStep": 0.023, "cosStep": 0.011 } }, "biasT": { "dtype": "float32", "shape": [1025], "data": { "kind": "fillFloat32", "scale": 0.05, "sinStep": 0.031, "cosStep": 0.007 } } }, "outputs": { "outputT": { "dtype": "float32", "shape": [3, 1025], "tolerance": 0.0005 }, "residualT": { "dtype": "float32", "shape": [3, 1025], "tolerance": 0.000002 } } }, { "name": "f32_hidden770_unaligned_beta_bias_scalar_subgroup", "attrs": { "epsilon": 0.00001 }, "inputs": { "inputT": { "dtype": "float32", "shape": [3, 770], "data": { "kind": "fillFloat32", "scale": 0.2, "sinStep": 0.017, "cosStep": 0.031 } }, "skipT": { "dtype": "float32", "shape": [3, 770], "data": { "kind": "fillFloat32", "scale": 0.2, "sinStep": 0.011, "cosStep": 0.023 } }, "gammaT": { "dtype": "float32", "shape": [770], "data": { "kind": "fillFloat32", "scale": 0.1, "offset": 1.0, "sinStep": 0.007, "cosStep": 0.041 } }, "betaT": { "dtype": "float32", "shape": [770], "data": { "kind": "fillFloat32", "scale": 0.1, "sinStep": 0.019, "cosStep": 0.013 } }, "biasT": { "dtype": "float32", "shape": [770], "data": { "kind": "fillFloat32", "scale": 0.05, "sinStep": 0.031, "cosStep": 0.007 } } }, "outputs": { "outputT": { "dtype": "float32", "shape": [3, 770], "tolerance": 0.0002 }, "residualT": { "dtype": "float32", "shape": [3, 770], "tolerance": 0.000002 } } }, { "name": "f32_rows1_hidden4096_decode", "attrs": { "epsilon": 0.00001 }, "inputs": { "inputT": { "dtype": "float32", "shape": [1, 4096], "data": { "kind": "fillFloat32", "scale": 0.2, "sinStep": 0.017, "cosStep": 0.031 } }, "skipT": { "dtype": "float32", "shape": [1, 4096], "data": { "kind": "fillFloat32", "scale": 0.2, "sinStep": 0.011, "cosStep": 0.023 } }, "gammaT": { "dtype": "float32", "shape": [4096], "data": { "kind": "fillFloat32", "scale": 0.1, "offset": 1.0, "sinStep": 0.007, "cosStep": 0.041 } }, "betaT": { "dtype": "float32", "shape": [4096], "data": { "kind": "fillFloat32", "scale": 0.1, "sinStep": 0.019, "cosStep": 0.013 } } }, "outputs": { "outputT": { "dtype": "float32", "shape": [1, 4096], "tolerance": 0.0005 }, "residualT": { "dtype": "float32", "shape": [1, 4096], "tolerance": 0.000002 } } }, { "name": "empty_tokens_bias_residual_twopass", "attrs": { "epsilon": 1e-12 }, "inputs": { "inputT": { "dtype": "float32", "shape": [0, 4], "data": { "kind": "values", "values": [] } }, "skipT": { "dtype": "float32", "shape": [0, 4], "data": { "kind": "values", "values": [] } }, "gammaT": { "dtype": "float32", "shape": [4], "data": { "kind": "values", "values": [0.3, 0.2, 4.0, 2.2] } }, "betaT": { "dtype": "float32", "shape": [4], "data": { "kind": "values", "values": [0.2, 0.1, 0.4, 1.6] } }, "biasT": { "dtype": "float32", "shape": [4], "data": { "kind": "values", "values": [0.1, -0.1, 0.2, -0.2] } } }, "outputs": { "outputT": { "dtype": "float32", "shape": [0, 4], "tolerance": 0 }, "residualT": { "dtype": "float32", "shape": [0, 4], "tolerance": 0 } } }, { "name": "rows65537_hidden3_fold_lastrow_guard", "provenance": { "notes": "Pins the two-dimensional dispatch fold and final-row guard with 65,537 hidden-size-3 rows; near-constant rows exercise float32 one-pass variance while the residual sum remains exact." }, "attrs": { "epsilon": 0.00001 }, "inputs": { "inputT": { "dtype": "float32", "shape": [65537, 3], "data": { "kind": "fillFloat32", "scale": 0.2, "sinStep": 0.017, "cosStep": 0.031 } }, "skipT": { "dtype": "float32", "shape": [65537, 3], "data": { "kind": "fillFloat32", "scale": 0.2, "sinStep": 0.011, "cosStep": 0.023 } }, "gammaT": { "dtype": "float32", "shape": [3], "data": { "kind": "values", "values": [1.0, 2.0, -1.0] } }, "betaT": { "dtype": "float32", "shape": [3], "data": { "kind": "values", "values": [0.5, -0.25, 1.0] } } }, "outputs": { "outputT": { "dtype": "float32", "shape": [65537, 3], "tolerance": 0.0021 }, "residualT": { "dtype": "float32", "shape": [65537, 3], "tolerance": 0.000002 } } }, { "name": "hidden_size_one_bias_many_rows", "provenance": { "notes": "Many hidden-size-one rows exercise the variance-zero bias path at a compact scale." }, "attrs": { "epsilon": 0.00001 }, "inputs": { "inputT": { "dtype": "float32", "shape": [257, 1], "data": { "kind": "fillFloat32", "scale": 0.2, "sinStep": 0.017, "cosStep": 0.031 } }, "skipT": { "dtype": "float32", "shape": [257, 1], "data": { "kind": "fillFloat32", "scale": 0.2, "sinStep": 0.011, "cosStep": 0.023 } }, "gammaT": { "dtype": "float32", "shape": [1], "data": { "kind": "values", "values": [1.25] } }, "betaT": { "dtype": "float32", "shape": [1], "data": { "kind": "values", "values": [-0.5] } }, "biasT": { "dtype": "float32", "shape": [1], "data": { "kind": "values", "values": [0.125] } } }, "outputs": { "outputT": { "dtype": "float32", "shape": [257, 1], "tolerance": 0.000001 }, "residualT": { "dtype": "float32", "shape": [257, 1], "tolerance": 0.000001 } } }, { "name": "rank3_no_bias_residual", "provenance": { "notes": "Rank-3 activation shape carrying the residual output, the form ONNX Runtime's transformer fusion emits when the pre-normalization sum feeds the next block." }, "attrs": { "epsilon": 0.00001 }, "inputs": { "inputT": { "dtype": "float32", "shape": [1, 4, 768], "data": { "kind": "fillFloat32", "scale": 0.2, "sinStep": 0.017, "cosStep": 0.031 } }, "skipT": { "dtype": "float32", "shape": [1, 4, 768], "data": { "kind": "fillFloat32", "scale": 0.2, "sinStep": 0.011, "cosStep": 0.023 } }, "gammaT": { "dtype": "float32", "shape": [768], "data": { "kind": "fillFloat32", "scale": 0.1, "offset": 1.0, "sinStep": 0.007, "cosStep": 0.041 } }, "betaT": { "dtype": "float32", "shape": [768], "data": { "kind": "fillFloat32", "scale": 0.1, "sinStep": 0.019, "cosStep": 0.013 } } }, "outputs": { "outputT": { "dtype": "float32", "shape": [1, 4, 768], "tolerance": 0.0002 }, "residualT": { "dtype": "float32", "shape": [1, 4, 768], "tolerance": 0.000002 } } }, { "name": "rank3_beta_no_bias_output_only_hidden6_unaligned_row", "provenance": { "notes": "Rank-3 at a hidden size that is not a multiple of four, so the scalar row route serves it." }, "attrs": { "epsilon": 0.00001 }, "inputs": { "inputT": { "dtype": "float32", "shape": [1, 2, 6], "data": { "kind": "fillFloat32", "sinStep": 0.19, "cosStep": 0.37 } }, "skipT": { "dtype": "float32", "shape": [1, 2, 6], "data": { "kind": "fillFloat32", "sinStep": 0.23, "cosStep": 0.43 } }, "gammaT": { "dtype": "float32", "shape": [6], "data": { "kind": "fillFloat32", "sinStep": 0.07, "cosStep": 0.41, "scale": 0.2 } }, "betaT": { "dtype": "float32", "shape": [6], "data": { "kind": "fillFloat32", "sinStep": 0.13, "cosStep": 0.29, "scale": 0.1 } } }, "outputs": { "outputT": { "dtype": "float32", "shape": [1, 2, 6], "tolerance": 0.00002 } } }, { "name": "f16_rank3_beta_bias_output_only", "provenance": { "notes": "Half-precision beta and bias at a rank-3 activation shape, the ordinary on-device inference form." }, "requires": { "features": ["shader-f16"] }, "attrs": { "epsilon": 0.00001 }, "inputs": { "inputT": { "dtype": "float16", "shape": [1, 3, 8], "data": { "kind": "fillFloat32", "sinStep": 0.17, "cosStep": 0.31 } }, "skipT": { "dtype": "float16", "shape": [1, 3, 8], "data": { "kind": "fillFloat32", "sinStep": 0.11, "cosStep": 0.23 } }, "gammaT": { "dtype": "float16", "shape": [8], "data": { "kind": "fillFloat32", "sinStep": 0.07, "cosStep": 0.41, "scale": 0.2 } }, "betaT": { "dtype": "float16", "shape": [8], "data": { "kind": "fillFloat32", "sinStep": 0.13, "cosStep": 0.29, "scale": 0.1 } }, "biasT": { "dtype": "float16", "shape": [8], "data": { "kind": "fillFloat32", "sinStep": 0.05, "cosStep": 0.47, "scale": 0.3 } } }, "outputs": { "outputT": { "dtype": "float16", "shape": [1, 3, 8], "tolerance": 0.01 } } }, { "name": "f16_beta_bias_output_only_hidden6_unaligned_row", "provenance": { "notes": "Half precision at an unaligned hidden size, which the scalar row route serves." }, "requires": { "features": ["shader-f16"] }, "attrs": { "epsilon": 0.00001 }, "inputs": { "inputT": { "dtype": "float16", "shape": [2, 6], "data": { "kind": "fillFloat32", "sinStep": 0.19, "cosStep": 0.37 } }, "skipT": { "dtype": "float16", "shape": [2, 6], "data": { "kind": "fillFloat32", "sinStep": 0.23, "cosStep": 0.43 } }, "gammaT": { "dtype": "float16", "shape": [6], "data": { "kind": "fillFloat32", "sinStep": 0.07, "cosStep": 0.41, "scale": 0.2 } }, "betaT": { "dtype": "float16", "shape": [6], "data": { "kind": "fillFloat32", "sinStep": 0.13, "cosStep": 0.29, "scale": 0.1 } }, "biasT": { "dtype": "float16", "shape": [6], "data": { "kind": "fillFloat32", "sinStep": 0.05, "cosStep": 0.47, "scale": 0.3 } } }, "outputs": { "outputT": { "dtype": "float16", "shape": [2, 6], "tolerance": 0.01 } } }, { "name": "f16_beta_no_bias_output_only", "provenance": { "notes": "Half precision with beta and no bias." }, "requires": { "features": ["shader-f16"] }, "attrs": { "epsilon": 0.00001 }, "inputs": { "inputT": { "dtype": "float16", "shape": [3, 8], "data": { "kind": "fillFloat32", "sinStep": 0.17, "cosStep": 0.31 } }, "skipT": { "dtype": "float16", "shape": [3, 8], "data": { "kind": "fillFloat32", "sinStep": 0.11, "cosStep": 0.23 } }, "gammaT": { "dtype": "float16", "shape": [8], "data": { "kind": "fillFloat32", "sinStep": 0.07, "cosStep": 0.41, "scale": 0.2 } }, "betaT": { "dtype": "float16", "shape": [8], "data": { "kind": "fillFloat32", "sinStep": 0.13, "cosStep": 0.29, "scale": 0.1 } } }, "outputs": { "outputT": { "dtype": "float16", "shape": [3, 8], "tolerance": 0.01 } } }, { "name": "f16_beta_no_bias_output_only_hidden6_unaligned_row", "provenance": { "notes": "Half precision with beta, no bias, at an unaligned hidden size." }, "requires": { "features": ["shader-f16"] }, "attrs": { "epsilon": 0.00001 }, "inputs": { "inputT": { "dtype": "float16", "shape": [2, 6], "data": { "kind": "fillFloat32", "sinStep": 0.19, "cosStep": 0.37 } }, "skipT": { "dtype": "float16", "shape": [2, 6], "data": { "kind": "fillFloat32", "sinStep": 0.23, "cosStep": 0.43 } }, "gammaT": { "dtype": "float16", "shape": [6], "data": { "kind": "fillFloat32", "sinStep": 0.07, "cosStep": 0.41, "scale": 0.2 } }, "betaT": { "dtype": "float16", "shape": [6], "data": { "kind": "fillFloat32", "sinStep": 0.13, "cosStep": 0.29, "scale": 0.1 } } }, "outputs": { "outputT": { "dtype": "float16", "shape": [2, 6], "tolerance": 0.01 } } }, { "name": "f16_no_beta_output_only_hidden6_unaligned_row", "provenance": { "notes": "Half precision with neither beta nor bias, at an unaligned hidden size." }, "requires": { "features": ["shader-f16"] }, "attrs": { "epsilon": 0.00001 }, "inputs": { "inputT": { "dtype": "float16", "shape": [2, 6], "data": { "kind": "fillFloat32", "sinStep": 0.17, "cosStep": 0.31 } }, "skipT": { "dtype": "float16", "shape": [2, 6], "data": { "kind": "fillFloat32", "sinStep": 0.11, "cosStep": 0.23 } }, "gammaT": { "dtype": "float16", "shape": [6], "data": { "kind": "fillFloat32", "sinStep": 0.07, "cosStep": 0.41, "scale": 0.2 } } }, "outputs": { "outputT": { "dtype": "float16", "shape": [2, 6], "tolerance": 0.002 } } }, { "name": "f16_no_beta_output_only", "provenance": { "notes": "Half precision with neither beta nor bias at an aligned hidden size, which is the vec4 arm of that pair." }, "requires": { "features": ["shader-f16"] }, "attrs": { "epsilon": 0.00001 }, "inputs": { "inputT": { "dtype": "float16", "shape": [3, 8], "data": { "kind": "fillFloat32", "sinStep": 0.17, "cosStep": 0.31 } }, "skipT": { "dtype": "float16", "shape": [3, 8], "data": { "kind": "fillFloat32", "sinStep": 0.11, "cosStep": 0.23 } }, "gammaT": { "dtype": "float16", "shape": [8], "data": { "kind": "fillFloat32", "sinStep": 0.07, "cosStep": 0.41, "scale": 0.2 } } }, "outputs": { "outputT": { "dtype": "float16", "shape": [3, 8], "tolerance": 0.002 } } } ] }