File size: 10,501 Bytes
aeefdfd
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3658569
aeefdfd
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3658569
aeefdfd
 
 
 
3658569
aeefdfd
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3658569
aeefdfd
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3658569
 
 
 
aeefdfd
 
 
 
 
 
 
 
 
 
3658569
 
 
 
aeefdfd
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
{
  "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": "On the scalar path, subnormal inputs produce 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 package's f32 large-argument range reduction differs from the correctly rounded expected value by about one ULP. More accurate range reduction or software-extended precision could close the 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 and expose a one-ULP float32 accuracy gap."
      },
      "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": "Near odd multiples of pi/2, tan(x) is highly sensitive to argument error. These moderate arguments stay below the 1e4 range-reduction threshold and exercise the direct tan implementation with an appropriate relative tolerance."
      },
      "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": 0.05, "relTolerance": 0.01 } },
      "provenance": {
        "notes": "Float16 arguments at the top of the range force argument reduction in float32; the tolerance is scaled to the result magnitude with headroom over the reduction residue."
      }
    },
    {
      "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": 0.05, "relTolerance": 0.01 } },
      "provenance": {
        "notes": "Large float16 arguments on the vec4 lane path force argument reduction in float32; the tolerance is scaled to the result magnitude with headroom over the reduction residue."
      }
    }
  ]
}