File size: 21,405 Bytes
d50509e
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
{
  "op": "ai.onnx.Compress",
  "cases": [
    {
      "name": "int16_axis_copy_boundaries",
      "attrs": { "axis": 0 },
      "inputs": {
        "input": {
          "dtype": "int16",
          "shape": [4, 2],
          "data": { "kind": "values", "values": [-32768, 32767, -1, 0, 1, -2, 32766, -32767] }
        },
        "condition": { "dtype": "bool", "shape": [4], "data": { "kind": "values", "values": [1, 0, 1, 1] } }
      },
      "outputs": {
        "output": {
          "dtype": "int16",
          "shape": [3, 2],
          "tolerance": 0,
          "data": { "kind": "values", "values": [-32768, 32767, 1, -2, 32766, -32767] }
        }
      }
    },
    {
      "name": "flatten_vit_token_prune_parallel_threshold",
      "inputs": {
        "input": { "dtype": "float32", "shape": [768], "data": { "kind": "linspace", "start": -1.0, "end": 1.0 } },
        "condition": { "dtype": "bool", "shape": [768], "data": { "kind": "cycle", "values": [1, 0] } }
      },
      "outputs": { "output": { "dtype": "float32", "shape": [384], "tolerance": 0 } }
    },
    {
      "name": "axis0_f32",
      "attrs": { "axis": 0 },
      "inputs": {
        "input": {
          "dtype": "float32",
          "shape": [3, 2],
          "data": { "kind": "values", "values": [1.0, 2.0, 3.0, 4.0, 5.0, 6.0] }
        },
        "condition": { "dtype": "bool", "shape": [3], "data": { "kind": "values", "values": [0, 1, 1] } }
      },
      "outputs": { "output": { "dtype": "float32", "shape": [2, 2] } },
      "provenance": {
        "source": "onnxruntime/test/providers/cpu/tensor/compress_op.test.cc",
        "test": "CompressTest.Compress0"
      }
    },
    {
      "name": "axis1_f32",
      "attrs": { "axis": 1 },
      "inputs": {
        "input": {
          "dtype": "float32",
          "shape": [3, 2],
          "data": { "kind": "values", "values": [1.0, 2.0, 3.0, 4.0, 5.0, 6.0] }
        },
        "condition": { "dtype": "bool", "shape": [2], "data": { "kind": "values", "values": [0, 1] } }
      },
      "outputs": { "output": { "dtype": "float32", "shape": [3, 1] } },
      "provenance": {
        "source": "onnxruntime/test/providers/cpu/tensor/compress_op.test.cc",
        "test": "CompressTest.Compress1"
      }
    },
    {
      "name": "axis1_3d_f32",
      "attrs": { "axis": 1 },
      "inputs": {
        "input": {
          "dtype": "float32",
          "shape": [2, 2, 3],
          "data": { "kind": "values", "values": [1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0, 10.0, 11.0, 12.0] }
        },
        "condition": { "dtype": "bool", "shape": [2], "data": { "kind": "values", "values": [0, 1] } }
      },
      "outputs": { "output": { "dtype": "float32", "shape": [2, 1, 3] } },
      "provenance": {
        "source": "onnxruntime/test/providers/cpu/tensor/compress_op.test.cc",
        "test": "CompressTest.Compress_3dims"
      }
    },
    {
      "name": "ort_axis1_long_condition_parallel_scan_compaction",
      "provenance": {
        "source": "onnxruntime/test/providers/cpu/tensor/compress_op.test.cc",
        "test": "CompressTest.Compress_3dims",
        "notes": "Extends ORT's axis=1 3D case to a long axis with a nontrivial true/false pattern so the WebGPU axis parallel prefix-scan path is exercised."
      },
      "attrs": { "axis": 1 },
      "inputs": {
        "input": {
          "dtype": "float32",
          "shape": [2, 1030, 2],
          "data": { "kind": "linspace", "start": 0.0, "end": 4119.0 }
        },
        "condition": { "dtype": "bool", "shape": [1030], "data": { "kind": "cycle", "values": [1, 0, 1, 1] } }
      },
      "outputs": { "output": { "dtype": "float32", "shape": [2, 772, 2], "tolerance": 0 } }
    },
    {
      "name": "all_false_zero_axis_dim",
      "attrs": { "axis": 1 },
      "inputs": {
        "input": {
          "dtype": "float32",
          "shape": [2, 2, 3],
          "data": { "kind": "values", "values": [1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0, 10.0, 11.0, 12.0] }
        },
        "condition": { "dtype": "bool", "shape": [2], "data": { "kind": "values", "values": [0, 0] } }
      },
      "outputs": { "output": { "dtype": "float32", "shape": [2, 0, 3] } },
      "provenance": {
        "source": "onnxruntime/test/providers/cpu/tensor/compress_op.test.cc",
        "test": "CompressTest.Compress_condition_all_false"
      }
    },
    {
      "name": "extra_condition_ignored",
      "attrs": { "axis": 1 },
      "inputs": {
        "input": {
          "dtype": "float32",
          "shape": [2, 2, 3],
          "data": { "kind": "values", "values": [1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0, 10.0, 11.0, 12.0] }
        },
        "condition": { "dtype": "bool", "shape": [3], "data": { "kind": "values", "values": [0, 1, 1] } }
      },
      "outputs": { "output": { "dtype": "float32", "shape": [2, 1, 3] } },
      "provenance": {
        "source": "onnxruntime/test/providers/cpu/tensor/compress_op.test.cc",
        "test": "CompressTest.Compress_3dims_has_extra_condition"
      }
    },
    {
      "name": "short_condition_ignores_extra_input_axis_values",
      "attrs": { "axis": 1 },
      "inputs": {
        "input": {
          "dtype": "float32",
          "shape": [2, 3, 3],
          "data": {
            "kind": "values",
            "values": [1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0, 10.0, 11.0, 12.0, 13.0, 14.0, 15.0, 16.0, 17.0, 18.0]
          }
        },
        "condition": { "dtype": "bool", "shape": [2], "data": { "kind": "values", "values": [0, 1] } }
      },
      "outputs": { "output": { "dtype": "float32", "shape": [2, 1, 3] } },
      "provenance": {
        "source": "onnxruntime/test/providers/cpu/tensor/compress_op.test.cc",
        "test": "CompressTest.Compress_3dims_has_extra_input"
      }
    },
    {
      "name": "default_axis_flatten_f32",
      "inputs": {
        "input": {
          "dtype": "float32",
          "shape": [3, 2],
          "data": { "kind": "values", "values": [1.0, 2.0, 3.0, 4.0, 5.0, 6.0] }
        },
        "condition": { "dtype": "bool", "shape": [5], "data": { "kind": "values", "values": [0, 1, 0, 0, 1] } }
      },
      "outputs": { "output": { "dtype": "float32", "shape": [2] } },
      "provenance": {
        "source": "onnxruntime/test/providers/cpu/tensor/compress_op.test.cc",
        "test": "CompressTest.Compress_default_axis"
      }
    },
    {
      "name": "ort_default_axis_all_true_parallel_scan_over_1024",
      "provenance": {
        "source": "onnxruntime/test/providers/cpu/tensor/compress_op.test.cc",
        "test": "CompressTest.Compress_default_axis_issue_9247_cumulative_sum_overflow",
        "notes": "Extends ORT's >127-element cumulative-sum regression above 1024 elements so the WebGPU parallel prefix-scan path is exercised."
      },
      "inputs": {
        "input": { "dtype": "float32", "shape": [23, 50], "data": { "kind": "linspace", "start": 0.0, "end": 1149.0 } },
        "condition": { "dtype": "bool", "shape": [1150], "data": { "kind": "constant", "value": 1 } }
      },
      "outputs": { "output": { "dtype": "float32", "shape": [1150], "tolerance": 0 } }
    },
    {
      "name": "negative_axis_f32",
      "attrs": { "axis": -2 },
      "inputs": {
        "input": {
          "dtype": "float32",
          "shape": [2, 2, 3],
          "data": { "kind": "values", "values": [1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0, 10.0, 11.0, 12.0] }
        },
        "condition": { "dtype": "bool", "shape": [2], "data": { "kind": "values", "values": [0, 1] } }
      },
      "outputs": { "output": { "dtype": "float32", "shape": [2, 1, 3] } },
      "provenance": {
        "source": "onnxruntime/test/providers/cpu/tensor/compress_op.test.cc",
        "test": "CompressTest.Compress_3dims_neg_axis"
      }
    },
    {
      "name": "axis0_uint32",
      "attrs": { "axis": 0 },
      "inputs": {
        "input": { "dtype": "uint32", "shape": [3, 2], "data": { "kind": "values", "values": [1, 2, 3, 4, 5, 6] } },
        "condition": { "dtype": "bool", "shape": [3], "data": { "kind": "values", "values": [1, 0, 1] } }
      },
      "outputs": { "output": { "dtype": "uint32", "shape": [2, 2] } }
    },
    {
      "name": "axis1_f16",
      "attrs": { "axis": 1 },
      "inputs": {
        "input": {
          "dtype": "float16",
          "shape": [2, 3],
          "data": { "kind": "values", "values": [1.0, 2.0, 3.0, 4.0, 5.0, 6.0] }
        },
        "condition": { "dtype": "bool", "shape": [3], "data": { "kind": "values", "values": [1, 0, 1] } }
      },
      "outputs": { "output": { "dtype": "float16", "shape": [2, 2] } },
      "tolerance": 0.001
    },
    {
      "name": "ort_default_axis_issue_9247_all_true_150",
      "inputs": {
        "input": {
          "dtype": "float32",
          "shape": [2, 75],
          "data": { "kind": "fillFloat32", "sinStep": 0.07, "cosStep": 0.11, "scale": 1.0 }
        },
        "condition": { "dtype": "bool", "shape": [150], "data": { "kind": "constant", "value": 1 } }
      },
      "outputs": { "output": { "dtype": "float32", "shape": [150], "tolerance": 0 } },
      "provenance": {
        "source": "onnxruntime/test/providers/cpu/tensor/compress_op.test.cc",
        "test": "CompressTest.Compress_default_axis_issue_9247_cumulative_sum_overflow"
      }
    },
    {
      "name": "onnx_backend_compress_negative_axis",
      "attrs": { "axis": -1 },
      "inputs": {
        "input": {
          "dtype": "float32",
          "shape": [3, 2],
          "data": { "kind": "values", "values": [1.0, 2.0, 3.0, 4.0, 5.0, 6.0] }
        },
        "condition": { "dtype": "bool", "shape": [2], "data": { "kind": "values", "values": [0, 1] } }
      },
      "outputs": { "output": { "dtype": "float32", "shape": [3, 1] } },
      "provenance": { "source": "cmake/external/onnx/onnx/backend/test/data/node/test_compress_negative_axis" }
    },
    {
      "name": "parallel_flatten_exact_output_f32",
      "provenance": {
        "notes": "262144 flattened elements with an alternating condition exercise the three-pass scan with the exact 131072-element ONNX output extent."
      },
      "inputs": {
        "input": {
          "dtype": "float32",
          "shape": [512, 512],
          "data": { "kind": "fillFloat32", "sinStep": 0.13, "cosStep": 0.29 }
        },
        "condition": { "dtype": "bool", "shape": [262144], "data": { "kind": "cycle", "values": [1, 0] } }
      },
      "outputs": { "output": { "dtype": "float32", "shape": [131072], "tolerance": 0 } }
    },
    {
      "name": "parallel_flatten_condition_shorter_u32",
      "provenance": {
        "notes": "Condition shorter than the flattened input: the scan limit is min(inputCount, conditionCount) = 3000 and trailing input elements are dropped."
      },
      "inputs": {
        "input": { "dtype": "uint32", "shape": [8192], "data": { "kind": "cycle", "values": [5, 7, 11, 2, 9] } },
        "condition": { "dtype": "bool", "shape": [3000], "data": { "kind": "cycle", "values": [0, 1] } }
      },
      "outputs": { "output": { "dtype": "uint32", "shape": [1500] } }
    },
    {
      "name": "parallel_axis1_3d_f32",
      "attrs": { "axis": 1 },
      "provenance": {
        "notes": "Middle-axis parallel compress: the scan runs over 300 axis flags and the scatter covers all 76800 elements in parallel."
      },
      "inputs": {
        "input": {
          "dtype": "float32",
          "shape": [64, 300, 4],
          "data": { "kind": "fillFloat32", "sinStep": 0.11, "cosStep": 0.23 }
        },
        "condition": { "dtype": "bool", "shape": [300], "data": { "kind": "cycle", "values": [1, 0, 1] } }
      },
      "outputs": { "output": { "dtype": "float32", "shape": [64, 200, 4], "tolerance": 0 } }
    },
    {
      "name": "parallel_axis_negative_f16",
      "attrs": { "axis": -2 },
      "provenance": {
        "notes": "Negative axis + f16 payload + condition shorter than the axis (1024 < 2048); axis positions past the condition are dropped."
      },
      "inputs": {
        "input": {
          "dtype": "float16",
          "shape": [2, 2048, 4],
          "data": { "kind": "cycle", "values": [0.5, 1.5, -2.0, 3.0] }
        },
        "condition": { "dtype": "bool", "shape": [1024], "data": { "kind": "cycle", "values": [1, 0, 0, 1] } }
      },
      "outputs": { "output": { "dtype": "float16", "shape": [2, 512, 4], "tolerance": 0 } }
    },
    {
      "name": "parallel_axis0_condition_extra_f32",
      "attrs": { "axis": 0 },
      "provenance": {
        "notes": "Condition longer than the axis (2000 > 1500); extra condition entries are ignored. 900 of 1500 rows selected."
      },
      "inputs": {
        "input": {
          "dtype": "float32",
          "shape": [1500, 8],
          "data": { "kind": "fillFloat32", "sinStep": 0.19, "cosStep": 0.41 }
        },
        "condition": { "dtype": "bool", "shape": [2000], "data": { "kind": "cycle", "values": [1, 0, 1, 0, 1] } }
      },
      "outputs": { "output": { "dtype": "float32", "shape": [900, 8], "tolerance": 0 } }
    },
    {
      "name": "ort_bool_exact_condition_default_axis",
      "provenance": {
        "source": "onnxruntime/test/providers/cpu/tensor/compress_op.test.cc",
        "test": "CompressTest.Compress_default_axis"
      },
      "inputs": {
        "input": {
          "dtype": "float32",
          "shape": [3, 2],
          "data": { "kind": "values", "values": [1.0, 2.0, 3.0, 4.0, 5.0, 6.0] }
        },
        "condition": { "dtype": "bool", "shape": [5], "data": { "kind": "values", "values": [0, 1, 0, 0, 1] } }
      },
      "outputs": { "output": { "dtype": "float32", "shape": [2], "tolerance": 0 } }
    },
    {
      "name": "ort_bool_exact_condition_negative_axis",
      "provenance": {
        "source": "onnxruntime/test/providers/cpu/tensor/compress_op.test.cc",
        "test": "CompressTest.Compress_3dims_neg_axis",
        "notes": "Exact ORT negative-axis fixture with an ONNX bool condition."
      },
      "attrs": { "axis": -2 },
      "inputs": {
        "input": {
          "dtype": "float32",
          "shape": [2, 2, 3],
          "data": { "kind": "values", "values": [1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0, 10.0, 11.0, 12.0] }
        },
        "condition": { "dtype": "bool", "shape": [2], "data": { "kind": "values", "values": [0, 1] } }
      },
      "outputs": { "output": { "dtype": "float32", "shape": [2, 1, 3], "tolerance": 0 } }
    },
    {
      "name": "bool_payload_axis1",
      "provenance": {
        "source": "onnxruntime/test/providers/cpu/tensor/compress_op.test.cc",
        "test": "CompressTest.Compress1",
        "notes": "Same axis=1 compaction pattern as ORT's fixture, using ONNX-valid bool payload tensors."
      },
      "attrs": { "axis": 1 },
      "inputs": {
        "input": { "dtype": "bool", "shape": [2, 3], "data": { "kind": "values", "values": [1, 0, 1, 0, 1, 0] } },
        "condition": { "dtype": "bool", "shape": [3], "data": { "kind": "values", "values": [1, 0, 1] } }
      },
      "outputs": { "output": { "dtype": "bool", "shape": [2, 2], "tolerance": 0 } }
    },
    {
      "name": "bool_payload_axis1_all_false_zero_dim",
      "provenance": {
        "source": "onnxruntime/test/providers/cpu/tensor/compress_op.test.cc",
        "test": "CompressTest.Compress_condition_all_false",
        "notes": "All-false axis condition from ORT's zero-output case, using ONNX-valid bool payload tensors."
      },
      "attrs": { "axis": 1 },
      "inputs": {
        "input": {
          "dtype": "bool",
          "shape": [2, 2, 3],
          "data": { "kind": "values", "values": [1, 0, 1, 0, 1, 0, 0, 1, 0, 1, 0, 1] }
        },
        "condition": { "dtype": "bool", "shape": [2], "data": { "kind": "values", "values": [0, 0] } }
      },
      "outputs": { "output": { "dtype": "bool", "shape": [2, 0, 3], "tolerance": 0 } }
    },
    {
      "name": "int8_payload_axis1_parallel_scatter",
      "provenance": {
        "notes": "int8 activation payload through the axis_parallel_scan scatter (numel 12288 > 1024). int8 stores as i32 per scalarType, so the $scalar copy path is exercised with the full s8 range incl -128/127. ORT CPU runs Compress for int8."
      },
      "attrs": { "axis": 1 },
      "inputs": {
        "input": {
          "dtype": "int8",
          "shape": [16, 96, 8],
          "data": { "kind": "cycle", "values": [-128, -1, 0, 1, 127, 42, -77, 5] }
        },
        "condition": { "dtype": "bool", "shape": [96], "data": { "kind": "cycle", "values": [1, 0, 1] } }
      },
      "outputs": { "output": { "dtype": "int8", "shape": [16, 64, 8], "tolerance": 0 } }
    },
    {
      "name": "uint8_payload_flatten_parallel_scatter",
      "provenance": {
        "notes": "uint8 payload through flatten_parallel_scan (numel 4096 > 1024). uint8 widens to u32; covers the full 0..255 range through the 3-pass predicate-scan scatter. ORT CPU runs Compress for uint8."
      },
      "inputs": {
        "input": {
          "dtype": "uint8",
          "shape": [4096],
          "data": { "kind": "cycle", "values": [0, 1, 127, 128, 200, 255] }
        },
        "condition": { "dtype": "bool", "shape": [4096], "data": { "kind": "cycle", "values": [1, 1, 0, 1] } }
      },
      "outputs": { "output": { "dtype": "uint8", "shape": [3072], "tolerance": 0 } }
    },
    {
      "name": "int32_large_magnitude_above_2pow24_flatten",
      "provenance": {
        "notes": "int32 payload with magnitudes above 2^24 (incl ±2e9) through flatten_parallel_scan. Verifies the i32 scatter copy preserves values that would be lossy if carried through f32; comparison uses the int32 dtype so it stays exact. ORT CPU runs Compress for int32."
      },
      "inputs": {
        "input": {
          "dtype": "int32",
          "shape": [4096],
          "data": { "kind": "cycle", "values": [16777217, 33554433, -100000000, 2000000000, -2000000000, 16777216] }
        },
        "condition": { "dtype": "bool", "shape": [4096], "data": { "kind": "cycle", "values": [1, 0] } }
      },
      "outputs": { "output": { "dtype": "int32", "shape": [2048], "tolerance": 0 } }
    },
    {
      "name": "int32_axis0_serial_fallback",
      "provenance": {
        "notes": "int32 payload, axis=0, and a compact input select the serial axis kernel. Large-magnitude values confirm the single-lane copy is exact for i32; ORT CPU runs Compress for int32."
      },
      "attrs": { "axis": 0 },
      "inputs": {
        "input": {
          "dtype": "int32",
          "shape": [16, 8],
          "data": { "kind": "cycle", "values": [100000000, -100000000, 7, -7] }
        },
        "condition": { "dtype": "bool", "shape": [16], "data": { "kind": "cycle", "values": [1, 0, 1, 1] } }
      },
      "outputs": { "output": { "dtype": "int32", "shape": [12, 8], "tolerance": 0 } }
    },
    {
      "name": "bool_condition_axis0_selectable",
      "provenance": { "notes": "Bool condition selects the first and last slices along axis 0." },
      "attrs": { "axis": 0 },
      "inputs": {
        "input": {
          "dtype": "float32",
          "shape": [4, 3],
          "data": { "kind": "values", "values": [1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0, 10.0, 11.0, 12.0] }
        },
        "condition": { "dtype": "bool", "shape": [4], "data": { "kind": "values", "values": [1, 0, 1, 1] } }
      },
      "outputs": { "output": { "dtype": "float32", "shape": [3, 3], "tolerance": 0 } }
    },
    {
      "name": "bool_payload_flatten_parallel_scan_2048",
      "inputs": {
        "input": { "dtype": "bool", "shape": [64, 32], "data": { "kind": "cycle", "values": [1, 0, 1, 1, 0, 1, 0, 0] } },
        "condition": { "dtype": "bool", "shape": [2048], "data": { "kind": "cycle", "values": [1, 0] } }
      },
      "outputs": { "output": { "dtype": "bool", "shape": [1024], "tolerance": 0 } }
    },
    {
      "name": "int32_axis_last_serial",
      "attrs": { "axis": -1 },
      "inputs": {
        "input": {
          "dtype": "int32",
          "shape": [3, 4],
          "data": { "kind": "values", "values": [10, 20, 30, 40, 50, 60, 70, 80, 90, 100, 110, 120] }
        },
        "condition": { "dtype": "bool", "shape": [4], "data": { "kind": "values", "values": [1, 0, 1, 0] } }
      },
      "outputs": { "output": { "dtype": "int32", "shape": [3, 2], "tolerance": 0 } }
    },
    {
      "name": "int8_axis1_parallel_condition_shorter_than_axisDim",
      "attrs": { "axis": 1 },
      "inputs": {
        "input": {
          "dtype": "int8",
          "shape": [4, 96, 4],
          "data": { "kind": "cycle", "values": [-128, -1, 0, 1, 127, 42, -77, 5] }
        },
        "condition": { "dtype": "bool", "shape": [32], "data": { "kind": "cycle", "values": [1, 0, 0, 1] } }
      },
      "outputs": { "output": { "dtype": "int8", "shape": [4, 16, 4], "tolerance": 0 } }
    },
    {
      "name": "rank8_axis1",
      "attrs": { "axis": 1 },
      "inputs": {
        "input": {
          "dtype": "float32",
          "shape": [2, 3, 1, 1, 1, 1, 2, 2],
          "data": { "kind": "linspace", "start": 1.0, "end": 48.0 }
        },
        "condition": { "dtype": "bool", "shape": [3], "data": { "kind": "values", "values": [1, 0, 1] } }
      },
      "outputs": { "output": { "dtype": "float32", "shape": [2, 2, 1, 1, 1, 1, 2, 2], "tolerance": 0 } }
    }
  ]
}