| { |
| "op": "ai.onnx.Tan", |
| "cases": [ |
| { |
| "name": "f32_values", |
| "inputs": { |
| "x": { |
| "dtype": "float32", |
| "shape": [6], |
| "data": { "kind": "values", "values": [-1.0, -0.5, 0.0, 0.5, 1.0, 1.25] } |
| } |
| }, |
| "outputs": { "y": { "dtype": "float32", "shape": [6], "tolerance": 0.00001 } } |
| }, |
| { |
| "name": "f32_subnormal_identity_tail_gpu_gap", |
| "skipGpu": { |
| "category": "permanent", |
| "reason": "Portable WGSL floating-point semantics do not guarantee preservation of the subnormal values required by this fixture. Backend evidence: WebGPU/Metal flushes subnormals to zero (f32 and f16); the kernel cannot preserve denormal inputs/outputs bit-exactly." |
| }, |
| "provenance": { |
| "source": "onnxruntime/test/providers/cpu/math/element_wise_ops_test.cc", |
| "test": "MathOpTest.Tan", |
| "notes": "For tiny finite inputs tan(x) rounds back to x in float32; zero-flushing erases the signed tail." |
| }, |
| "inputs": { |
| "x": { |
| "dtype": "float32", |
| "shape": [4], |
| "data": { "kind": "values", "values": [-1e-39, -1e-40, 1e-40, 1e-39] } |
| } |
| }, |
| "outputs": { "y": { "dtype": "float32", "shape": [4], "tolerance": 0 } } |
| }, |
| { |
| "name": "f32_subnormal_identity_tail_scalar_gpu_gap", |
| "skipGpu": { |
| "category": "permanent", |
| "reason": "Portable WGSL floating-point semantics do not guarantee preservation of the subnormal values required by this fixture. Backend evidence: WebGPU/Metal flushes subnormals to zero (f32 and f16); the kernel cannot preserve denormal inputs/outputs bit-exactly." |
| }, |
| "provenance": { |
| "source": "onnxruntime/test/providers/cpu/math/element_wise_ops_test.cc", |
| "test": "MathOpTest.Tan", |
| "notes": "Scalar-path companion: subnormal inputs are valid finite Tan outputs." |
| }, |
| "inputs": { |
| "x": { "dtype": "float32", "shape": [3], "data": { "kind": "values", "values": [-1e-40, 0.0, 1e-40] } } |
| }, |
| "outputs": { "y": { "dtype": "float32", "shape": [3], "tolerance": 0 } } |
| }, |
| { |
| "name": "f16_values", |
| "inputs": { |
| "x": { |
| "dtype": "float16", |
| "shape": [2, 3], |
| "data": { "kind": "values", "values": [-1.0, -0.5, 0.0, 0.5, 1.0, 1.25] } |
| } |
| }, |
| "outputs": { "y": { "dtype": "float16", "shape": [2, 3] } }, |
| "tolerance": 0.002 |
| }, |
| { |
| "name": "large_argument_range_reduction", |
| "inputs": { |
| "x": { |
| "dtype": "float32", |
| "shape": [8], |
| "data": { |
| "kind": "values", |
| "values": [1000000.0, 10000000.0, 10000000000000.0, 100000000000000000000.0, -1000000000000000.0, -123456.78, 314159.265, 2500000000.0] |
| } |
| } |
| }, |
| "outputs": { "y": { "dtype": "float32", "shape": [8], "tolerance": 0.001 } } |
| }, |
| { |
| "name": "f32_large_argument_range_reduction_accuracy_gpu_gap", |
| "skipGpu": { |
| "category": "todo", |
| "reason": "The current f32 large-argument range reduction differs from the CPU reference by about one ULP. A more accurate range reduction or software-extended precision could close this implementable gap." |
| }, |
| "provenance": { |
| "source": "onnxruntime/test/providers/cpu/math/element_wise_ops_test.cc", |
| "test": "MathOpTest.Tan", |
| "notes": "Finite large arguments stress Tan's range reduction; ORT and the CPU reference agree more tightly than the current WGSL builtin path on these float32 inputs." |
| }, |
| "inputs": { |
| "x": { |
| "dtype": "float32", |
| "shape": [8], |
| "data": { |
| "kind": "values", |
| "values": [1000000.0, 10000000.0, 10000000000000.0, 100000000000000000000.0, -1000000000000000.0, -123456.78, 314159.265, 2500000000.0] |
| } |
| } |
| }, |
| "outputs": { "y": { "dtype": "float32", "shape": [8], "tolerance": 0.000001, "relTolerance": 0 } } |
| }, |
| { |
| "name": "ort_float_large_arguments", |
| "provenance": { |
| "source": "onnxruntime/test/providers/cpu/math/element_wise_ops_test.cc", |
| "test": "MathOpTest.Tan" |
| }, |
| "inputs": { |
| "x": { |
| "dtype": "float32", |
| "shape": [5], |
| "data": { "kind": "values", "values": [-100.0, -50.0, 0.0, 50.0, 100.0] } |
| } |
| }, |
| "outputs": { "y": { "dtype": "float32", "shape": [5], "tolerance": 0.0001 } } |
| }, |
| { |
| "name": "ort_nonfinite_values", |
| "provenance": { |
| "source": "onnxruntime/test/providers/cpu/math/element_wise_ops_test.cc", |
| "test": "MathOpTest.Tan", |
| "notes": "Extends ORT's large-argument case with signed infinities and NaN." |
| }, |
| "inputs": { |
| "x": { |
| "dtype": "float32", |
| "shape": [7], |
| "data": { "kind": "values", "values": ["-Infinity", -1.0, 0.0, 0.0, 1.0, "Infinity", "NaN"] } |
| } |
| }, |
| "outputs": { "y": { "dtype": "float32", "shape": [7], "tolerance": 0.0001, "allowNaN": true } } |
| }, |
| { |
| "name": "onnx_backend_example", |
| "provenance": { "source": "cmake/external/onnx/onnx/backend/test/data/node/test_tan_example" }, |
| "inputs": { "x": { "dtype": "float32", "shape": [3], "data": { "kind": "values", "values": [-1.0, 0.0, 1.0] } } }, |
| "outputs": { "y": { "dtype": "float32", "shape": [3], "tolerance": 0.000001 } } |
| }, |
| { |
| "name": "onnx_backend_tan", |
| "provenance": { "source": "cmake/external/onnx/onnx/backend/test/data/node/test_tan" }, |
| "inputs": { |
| "x": { |
| "dtype": "float32", |
| "shape": [3, 4, 5], |
| "data": { |
| "kind": "values", |
| "values": [1.764052391052246, 0.40015721321105957, 0.978738009929657, 2.2408931255340576, 1.8675580024719238, -0.9772778749465942, 0.9500884413719177, -0.15135720372200012, -0.10321885347366333, 0.4105985164642334, 0.14404356479644775, 1.4542734622955322, 0.7610377073287964, 0.12167501449584961, 0.44386324286460876, 0.3336743414402008, 1.4940791130065918, -0.2051582634449005, 0.3130677044391632, -0.8540957570075989, -2.5529897212982178, 0.653618574142456, 0.8644362092018127, -0.7421650290489197, 2.269754648208618, -1.4543657302856445, 0.04575851559638977, -0.18718385696411133, 1.5327792167663574, 1.4693588018417358, 0.154947429895401, 0.37816253304481506, -0.8877857327461243, -1.980796456336975, -0.34791216254234314, 0.15634897351264954, 1.2302906513214111, 1.202379822731018, -0.38732680678367615, -0.302302747964859, -1.0485529899597168, -1.420017957687378, -1.7062702178955078, 1.950775384902954, -0.5096521973609924, -0.4380742907524109, -1.2527953386306763, 0.7774903774261475, -1.6138978004455566, -0.21274028718471527, -0.8954665660858154, 0.38690251111984253, -0.5108051300048828, -1.18063223361969, -0.02818222902715206, 0.4283318817615509, 0.06651721894741058, 0.30247190594673157, -0.6343221068382263, -0.3627411723136902] |
| } |
| } |
| }, |
| "outputs": { "y": { "dtype": "float32", "shape": [3, 4, 5], "tolerance": 0.00001 } } |
| }, |
| { |
| "name": "onnx_backend_tan_example", |
| "provenance": { "source": "cmake/external/onnx/onnx/backend/test/data/node/test_tan_example" }, |
| "inputs": { "x": { "dtype": "float32", "shape": [3], "data": { "kind": "values", "values": [-1.0, 0.0, 1.0] } } }, |
| "outputs": { "y": { "dtype": "float32", "shape": [3], "tolerance": 0.00001 } } |
| }, |
| { |
| "name": "vec4_f16_lanes", |
| "inputs": { |
| "x": { |
| "dtype": "float16", |
| "shape": [16], |
| "data": { |
| "kind": "values", |
| "values": [-1.2, -1.0, -0.75, -0.5, -0.25, -0.125, 0.0, 0.125, 0.25, 0.5, 0.75, 1.0, 1.2, 2.0, 3.0, 5.0] |
| } |
| } |
| }, |
| "outputs": { "y": { "dtype": "float16", "shape": [16], "tolerance": 0.001, "relTolerance": 0.005 } } |
| }, |
| { |
| "name": "vec4_f32_nonfinite", |
| "inputs": { |
| "x": { |
| "dtype": "float32", |
| "shape": [8], |
| "data": { "kind": "values", "values": ["-Infinity", -1.0, 0.0, 0.5, 1.0, "Infinity", "NaN", 2.0] } |
| } |
| }, |
| "outputs": { "y": { "dtype": "float32", "shape": [8], "tolerance": 0.000001, "allowNaN": true } } |
| }, |
| { |
| "name": "empty_input_zero_dim", |
| "inputs": { "x": { "dtype": "float32", "shape": [0], "data": { "kind": "values", "values": [] } } }, |
| "outputs": { "y": { "dtype": "float32", "shape": [0], "tolerance": 0 } } |
| }, |
| { |
| "name": "near_pole_argument_accuracy", |
| "provenance": { |
| "source": "ai.onnx.Tan reference (Math.tan)", |
| "notes": "Tan blows up near odd multiples of pi/2; tan(x) ~ 1/(pi/2 - x) so a tiny argument error is amplified into a large relative output error. These arguments are below the kernel's 1e4 range-reduction threshold and so route through the hardware tan() directly. Validates that the Metal hardware tan() matches the CPU Math.tan reference at moderate near-pole arguments (the riskier untested regime; pi/4 and pi anchors are well-conditioned)." |
| }, |
| "inputs": { |
| "x": { |
| "dtype": "float32", |
| "shape": [8], |
| "data": { |
| "kind": "values", |
| "values": [1.5706963267948966, 1.5697963267948967, -1.5706963267948966, 4.711388980384689, 0.7853981633974483, -0.7853981633974483, 3.1405926535897932, 0.0] |
| } |
| } |
| }, |
| "outputs": { "y": { "dtype": "float32", "shape": [8], "tolerance": 0.01, "relTolerance": 0.0005 } } |
| }, |
| { |
| "name": "f16_scalar_near_f16_max", |
| "inputs": { |
| "x": { |
| "dtype": "float16", |
| "shape": [7], |
| "data": { "kind": "values", "values": [65504.0, -65504.0, 20000.0, -20000.0, 50000.0, -50000.0, 0.0] } |
| } |
| }, |
| "outputs": { "y": { "dtype": "float16", "shape": [7], "tolerance": 0.5, "relTolerance": 0.01 } } |
| }, |
| { |
| "name": "f16_vec4_large_arg_above_reduce_threshold", |
| "inputs": { |
| "x": { |
| "dtype": "float16", |
| "shape": [8], |
| "data": { "kind": "values", "values": [20000.0, -20000.0, 40000.0, -40000.0, 0.5, -0.5, 1.0, -1.0] } |
| } |
| }, |
| "outputs": { "y": { "dtype": "float16", "shape": [8], "tolerance": 0.5, "relTolerance": 0.01 } } |
| } |
| ] |
| } |
|
|
