| { |
| "fixtureArrays": { |
| "onnx_backend_roialign_input_x": [0.27639999985694885, 0.7149999737739563, 0.19580000638961792, 0.3416000008583069, 0.46380001306533813, 0.02590000070631504, 0.2962999939918518, 0.6517999768257141, 0.48559999465942383, 0.7250000238418579, 0.963699996471405, 0.08950000256299973, 0.29190000891685486, 0.6753000020980835, 0.023399999365210533, 0.6132000088691711, 0.8084999918937683, 0.5324000120162964, 0.8992000222206116, 0.44670000672340393, 0.3264999985694885, 0.8478999733924866, 0.9697999954223633, 0.24709999561309814, 0.9336000084877014, 0.18780000507831573, 0.4765999913215637, 0.4307999908924103, 0.3400000035762787, 0.21619999408721924, 0.020600000396370888, 0.1720000058412552, 0.21549999713897705, 0.43939998745918274, 0.06530000269412994, 0.34060001373291016, 0.7724000215530396, 0.3921000063419342, 0.2540999948978424, 0.5799000263214111, 0.40619999170303345, 0.21940000355243683, 0.447299987077713, 0.46869999170303345, 0.7109000086784363, 0.932699978351593, 0.9815000295639038, 0.6320000290870667, 0.1728000044822693, 0.6118999719619751, 0.30970001220703125, 0.1282999962568283, 0.4984000027179718, 0.5067999958992004, 0.4278999865055084, 0.01730000041425228, 0.43880000710487366, 0.0430000014603138, 0.46709999442100525, 0.711899995803833, 0.10109999775886536, 0.8476999998092651, 0.4726000130176544, 0.1776999980211258, 0.9922999739646912, 0.4041999876499176, 0.18690000474452972, 0.7795000076293945, 0.9945999979972839, 0.9689000248908997, 0.13660000264644623, 0.3671000003814697, 0.7010999917984009, 0.6233999729156494, 0.9866999983787537, 0.5584999918937683, 0.6984999775886536, 0.5608999729156494, 0.8787999749183655, 0.9927999973297119, 0.5697000026702881, 0.8511000275611877, 0.6711000204086304, 0.9405999779701233, 0.8751000165939331, 0.7495999932289124, 0.16500000655651093, 0.10490000247955322, 0.1559000015258789, 0.2513999938964844, 0.701200008392334, 0.40560001134872437, 0.7878999710083008, 0.34610000252723694, 0.04149999842047691, 0.29980000853538513, 0.5094000101089478, 0.3727000057697296, 0.5482000112533569, 0.050200000405311584], |
| "avg_single_roi_f32_input_x": [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15], |
| "ort_f16_half_pixel_8x8_sampling2_input_x": [0, 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], |
| "onnx_backend_aligned_false_5x5_input_x": [0.2764, 0.715, 0.1958, 0.3416, 0.4638, 0.0259, 0.2963, 0.6518, 0.4856, 0.725, 0.9637, 0.0895, 0.2919, 0.6753, 0.0234, 0.6132, 0.8085, 0.5324, 0.8992, 0.4467, 0.3265, 0.8479, 0.9698, 0.2471, 0.9336, 0.1878, 0.4766, 0.4308, 0.34, 0.2162, 0.0206, 0.172, 0.2155, 0.4394, 0.0653, 0.3406, 0.7724, 0.3921, 0.2541, 0.5799, 0.4062, 0.2194, 0.4473, 0.4687, 0.7109, 0.9327, 0.9815, 0.632, 0.1728, 0.6119, 0.3097, 0.1283, 0.4984, 0.5068, 0.4279, 0.0173, 0.4388, 0.043, 0.4671, 0.7119, 0.1011, 0.8477, 0.4726, 0.1777, 0.9923, 0.4042, 0.1869, 0.7795, 0.9946, 0.9689, 0.1366, 0.3671, 0.7011, 0.6234, 0.9867, 0.5585, 0.6985, 0.5609, 0.8788, 0.9928, 0.5697, 0.8511, 0.6711, 0.9406, 0.8751, 0.7496, 0.165, 0.1049, 0.1559, 0.2514, 0.7012, 0.4056, 0.7879, 0.3461, 0.0415, 0.2998, 0.5094, 0.3727, 0.5482, 0.0502], |
| "ort_out_of_bounds_degenerate_rois_half_pixel_input_x": [0, 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], |
| "ort_out_of_bounds_degenerate_rois_half_pixel_input_rois": [7, 5, 7, 5, -15, -15, -15, -15, -10, 21, -10, 21, 13, 8, 13, 8, -14, 19, -14, 19] |
| }, |
| "cases": [ |
| { |
| "name": "onnx_default_adaptive_sampling_omitted", |
| "provenance": { |
| "source": "https://onnx.ai/onnx/operators/onnx__RoiAlign.html", |
| "test": "RoiAlign sampling_ratio default", |
| "notes": "Omitted sampling_ratio, mode, and spatial_scale exercise their ONNX defaults (0, avg, and 1). The adaptive grid produces 0.625 for this nonlinear feature map; a fixed one-sample grid would produce 2.5." |
| }, |
| "attrs": { "coordinate_transformation_mode": "half_pixel" }, |
| "inputs": { |
| "x": { |
| "dtype": "float32", |
| "shape": [1, 1, 4, 4], |
| "data": { |
| "kind": "values", |
| "values": [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 10.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0] |
| } |
| }, |
| "rois": { "dtype": "float32", "shape": [1, 4], "data": { "kind": "values", "values": [0.0, 0.0, 4.0, 4.0] } }, |
| "batch_indices": { "dtype": "uint32", "shape": [1], "data": { "kind": "values", "values": [0] } } |
| }, |
| "outputs": { |
| "y": { |
| "dtype": "float32", |
| "shape": [1, 1, 1, 1], |
| "tolerance": 0.000001, |
| "data": { "kind": "values", "values": [0.625] } |
| } |
| } |
| }, |
| { |
| "name": "avg_half_pixel_exact_subnormal_pixel_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: Metal flushes denormals to zero in the ALU; RoiAlign interpolation/pooling arithmetic flushes the subnormal pixel on GPU. Permanent FTZ limitation." |
| }, |
| "provenance": { |
| "notes": "A zero-area half-pixel ROI samples exactly at pixel (0, 0); the positive subnormal payload should survive avg mode." |
| }, |
| "attrs": { |
| "spatial_scale": 1, |
| "sampling_ratio": 1, |
| "mode": "avg", |
| "coordinate_transformation_mode": "half_pixel", |
| "output_height": 1, |
| "output_width": 1 |
| }, |
| "inputs": { |
| "x": { |
| "dtype": "float32", |
| "shape": [1, 1, 2, 2], |
| "data": { "kind": "values", "values": [1e-40, 0.0, 0.0, 0.0] } |
| }, |
| "rois": { "dtype": "float32", "shape": [1, 4], "data": { "kind": "values", "values": [0.5, 0.5, 0.5, 0.5] } }, |
| "batch_indices": { "dtype": "uint32", "shape": [1], "data": { "kind": "values", "values": [0] } } |
| }, |
| "outputs": { "y": { "dtype": "float32", "shape": [1, 1, 1, 1], "tolerance": 0 } } |
| }, |
| { |
| "name": "max_half_pixel_exact_subnormal_pixel_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: Metal flushes denormals to zero in the ALU; RoiAlign interpolation/pooling arithmetic flushes the subnormal pixel on GPU. Permanent FTZ limitation." |
| }, |
| "provenance": { |
| "notes": "Max mode must preserve the exact positive subnormal sample because it is greater than every zero-valued neighbor." |
| }, |
| "attrs": { |
| "spatial_scale": 1, |
| "sampling_ratio": 1, |
| "mode": "max", |
| "coordinate_transformation_mode": "half_pixel", |
| "output_height": 1, |
| "output_width": 1 |
| }, |
| "inputs": { |
| "x": { |
| "dtype": "float32", |
| "shape": [1, 1, 2, 2], |
| "data": { "kind": "values", "values": [1e-40, 0.0, 0.0, 0.0] } |
| }, |
| "rois": { "dtype": "float32", "shape": [1, 4], "data": { "kind": "values", "values": [0.5, 0.5, 0.5, 0.5] } }, |
| "batch_indices": { "dtype": "uint32", "shape": [1], "data": { "kind": "values", "values": [0] } } |
| }, |
| "outputs": { "y": { "dtype": "float32", "shape": [1, 1, 1, 1], "tolerance": 0 } } |
| }, |
| { |
| "name": "max_half_pixel_fractional_sample_matches_ort", |
| "provenance": { |
| "source": "onnxruntime/test/providers/cpu/object_detection/roialign_test.cc", |
| "test": "RoiAlignTest.MaxModePositive", |
| "notes": "Max mode takes the maximum weighted corner contribution at a fractional sample point; this pins the unintuitive fractional-sample behavior." |
| }, |
| "attrs": { |
| "spatial_scale": 1, |
| "sampling_ratio": 1, |
| "mode": "max", |
| "coordinate_transformation_mode": "half_pixel", |
| "output_height": 1, |
| "output_width": 1 |
| }, |
| "inputs": { |
| "x": { "dtype": "float32", "shape": [1, 1, 2, 2], "data": { "kind": "values", "values": [1.0, 2.0, 3.0, 4.0] } }, |
| "rois": { "dtype": "float32", "shape": [1, 4], "data": { "kind": "values", "values": [0.0, 0.0, 2.0, 2.0] } }, |
| "batch_indices": { "dtype": "uint32", "shape": [1], "data": { "kind": "values", "values": [0] } } |
| }, |
| "outputs": { |
| "y": { |
| "dtype": "float32", |
| "shape": [1, 1, 1, 1], |
| "tolerance": 0.000001, |
| "data": { "kind": "values", "values": [1.0] } |
| } |
| } |
| }, |
| { |
| "name": "dispatch_cliff_nchw_over_16m_elements", |
| "attrs": { "spatial_scale": 1, "sampling_ratio": 1, "mode": "avg", "output_height": 64, "output_width": 64 }, |
| "inputs": { |
| "x": { "dtype": "float32", "shape": [1, 256, 16, 16] }, |
| "rois": { "dtype": "float32", "shape": [16, 4], "data": { "kind": "cycle", "values": [0.0, 0.0, 15.0, 15.0] } }, |
| "batch_indices": { "dtype": "uint32", "shape": [16], "data": { "kind": "constant", "value": 0 } } |
| }, |
| "outputs": { "y": { "dtype": "float32", "shape": [16, 256, 64, 64], "tolerance": 0.001 } } |
| }, |
| { |
| "name": "avg_single_roi_f32", |
| "attrs": { "spatial_scale": 1, "sampling_ratio": 2, "mode": "avg", "output_height": 2, "output_width": 2 }, |
| "inputs": { |
| "x": { |
| "dtype": "float32", |
| "shape": [1, 1, 4, 4], |
| "data": { "kind": "values", "values": { "$ref": "#/fixtureArrays/avg_single_roi_f32_input_x" } } |
| }, |
| "rois": { "dtype": "float32", "shape": [1, 4], "data": { "kind": "values", "values": [0.0, 0.0, 3.0, 3.0] } }, |
| "batch_indices": { "dtype": "uint32", "shape": [1], "data": { "kind": "values", "values": [0] } } |
| }, |
| "outputs": { "y": { "dtype": "float32", "shape": [1, 1, 2, 2], "tolerance": 0.000001 } } |
| }, |
| { |
| "name": "max_two_rois_f16", |
| "attrs": { "spatial_scale": 1, "sampling_ratio": 1, "mode": "max", "output_height": 1, "output_width": 1 }, |
| "inputs": { |
| "x": { |
| "dtype": "float16", |
| "shape": [1, 1, 4, 4], |
| "data": { |
| "kind": "values", |
| "values": [1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0, 10.0, 11.0, 12.0, 13.0, 14.0, 15.0, 16.0] |
| } |
| }, |
| "rois": { |
| "dtype": "float16", |
| "shape": [2, 4], |
| "data": { "kind": "values", "values": [0.0, 0.0, 2.0, 2.0, 1.0, 1.0, 2.0, 2.0] } |
| }, |
| "batch_indices": { "dtype": "uint32", "shape": [2], "data": { "kind": "values", "values": [0, 0] } } |
| }, |
| "outputs": { "y": { "dtype": "float16", "shape": [2, 1, 1, 1], "tolerance": 0.002 } } |
| }, |
| { |
| "name": "ort_f16_half_pixel_8x8_sampling2", |
| "provenance": { |
| "source": "onnxruntime/test/providers/cpu/object_detection/roialign_test.cc", |
| "test": "RoiAlignTest.Float16_HalfPixel_Opset16" |
| }, |
| "attrs": { |
| "spatial_scale": 1, |
| "sampling_ratio": 2, |
| "mode": "avg", |
| "coordinate_transformation_mode": "half_pixel", |
| "output_height": 2, |
| "output_width": 2 |
| }, |
| "inputs": { |
| "x": { |
| "dtype": "float16", |
| "shape": [1, 1, 8, 8], |
| "data": { "kind": "values", "values": { "$ref": "#/fixtureArrays/ort_f16_half_pixel_8x8_sampling2_input_x" } } |
| }, |
| "rois": { "dtype": "float16", "shape": [1, 4], "data": { "kind": "values", "values": [0.0, 0.0, 7.0, 7.0] } }, |
| "batch_indices": { "dtype": "uint32", "shape": [1], "data": { "kind": "values", "values": [0] } } |
| }, |
| "outputs": { "y": { "dtype": "float16", "shape": [1, 1, 2, 2], "tolerance": 0.01 } } |
| }, |
| { |
| "name": "ort_f16_output_half_pixel_2x2_sampling1", |
| "provenance": { |
| "source": "onnxruntime/test/providers/cpu/object_detection/roialign_test.cc", |
| "test": "RoiAlignTest.Float16_Opset22" |
| }, |
| "attrs": { |
| "spatial_scale": 1, |
| "sampling_ratio": 1, |
| "mode": "avg", |
| "coordinate_transformation_mode": "output_half_pixel", |
| "output_height": 1, |
| "output_width": 1 |
| }, |
| "inputs": { |
| "x": { "dtype": "float16", "shape": [1, 1, 2, 2], "data": { "kind": "values", "values": [1.0, 2.0, 1.0, 1.0] } }, |
| "rois": { "dtype": "float16", "shape": [1, 4], "data": { "kind": "values", "values": [0.0, 0.0, 1.0, 1.0] } }, |
| "batch_indices": { "dtype": "uint32", "shape": [1], "data": { "kind": "values", "values": [0] } } |
| }, |
| "outputs": { "y": { "dtype": "float16", "shape": [1, 1, 1, 1], "tolerance": 0.01 } } |
| }, |
| { |
| "name": "half_pixel_small_roi_no_min_size_f32", |
| "attrs": { "spatial_scale": 1, "sampling_ratio": 1, "mode": "avg", "output_height": 1, "output_width": 1 }, |
| "inputs": { |
| "x": { |
| "dtype": "float32", |
| "shape": [1, 1, 4, 4], |
| "data": { "kind": "values", "values": { "$ref": "#/fixtureArrays/avg_single_roi_f32_input_x" } } |
| }, |
| "rois": { "dtype": "float32", "shape": [1, 4], "data": { "kind": "values", "values": [1.0, 1.0, 1.5, 1.5] } }, |
| "batch_indices": { "dtype": "uint32", "shape": [1], "data": { "kind": "values", "values": [0] } } |
| }, |
| "outputs": { "y": { "dtype": "float32", "shape": [1, 1, 1, 1], "tolerance": 0.000001 } } |
| }, |
| { |
| "name": "avg_output_half_pixel_mode", |
| "attrs": { |
| "spatial_scale": 1, |
| "sampling_ratio": 2, |
| "mode": "avg", |
| "coordinate_transformation_mode": "output_half_pixel", |
| "output_height": 2, |
| "output_width": 2 |
| }, |
| "inputs": { |
| "x": { |
| "dtype": "float32", |
| "shape": [1, 1, 4, 4], |
| "data": { "kind": "values", "values": { "$ref": "#/fixtureArrays/avg_single_roi_f32_input_x" } } |
| }, |
| "rois": { "dtype": "float32", "shape": [1, 4], "data": { "kind": "values", "values": [0.0, 0.0, 3.0, 3.0] } }, |
| "batch_indices": { "dtype": "uint32", "shape": [1], "data": { "kind": "values", "values": [0] } } |
| }, |
| "outputs": { "y": { "dtype": "float32", "shape": [1, 1, 2, 2], "tolerance": 0.000001 } } |
| }, |
| { |
| "name": "avg_two_batches_two_channels", |
| "attrs": { "spatial_scale": 1, "sampling_ratio": 1, "mode": "avg", "output_height": 1, "output_width": 1 }, |
| "inputs": { |
| "x": { |
| "dtype": "float32", |
| "shape": [2, 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, 20.0, 21.0, 22.0, 23.0, 24.0, 25.0, 26.0, 27.0, 28.0, 30.0, 31.0, 32.0, 33.0, 34.0, 35.0, 36.0, 37.0, 38.0] |
| } |
| }, |
| "rois": { |
| "dtype": "float32", |
| "shape": [2, 4], |
| "data": { "kind": "values", "values": [0.0, 0.0, 2.0, 2.0, 1.0, 1.0, 2.0, 2.0] } |
| }, |
| "batch_indices": { "dtype": "uint32", "shape": [2], "data": { "kind": "values", "values": [0, 1] } } |
| }, |
| "outputs": { "y": { "dtype": "float32", "shape": [2, 2, 1, 1], "tolerance": 0.000001 } } |
| }, |
| { |
| "name": "onnx_backend_aligned_false_5x5", |
| "provenance": { |
| "source": "cmake/external/onnx/onnx/backend/test/case/node/roialign.py", |
| "test": "test_roialign_aligned_false" |
| }, |
| "attrs": { |
| "spatial_scale": 1, |
| "sampling_ratio": 2, |
| "mode": "avg", |
| "coordinate_transformation_mode": "output_half_pixel", |
| "output_height": 5, |
| "output_width": 5 |
| }, |
| "inputs": { |
| "x": { |
| "dtype": "float32", |
| "shape": [1, 1, 10, 10], |
| "data": { "kind": "values", "values": { "$ref": "#/fixtureArrays/onnx_backend_aligned_false_5x5_input_x" } } |
| }, |
| "rois": { |
| "dtype": "float32", |
| "shape": [3, 4], |
| "data": { "kind": "values", "values": [0.0, 0.0, 9.0, 9.0, 0.0, 5.0, 4.0, 9.0, 5.0, 5.0, 9.0, 9.0] } |
| }, |
| "batch_indices": { "dtype": "uint32", "shape": [3], "data": { "kind": "values", "values": [0, 0, 0] } } |
| }, |
| "outputs": { "y": { "dtype": "float32", "shape": [3, 1, 5, 5], "tolerance": 0.0002 } } |
| }, |
| { |
| "name": "onnx_backend_aligned_true_5x5", |
| "provenance": { |
| "source": "cmake/external/onnx/onnx/backend/test/case/node/roialign.py", |
| "test": "test_roialign_aligned_true" |
| }, |
| "attrs": { |
| "spatial_scale": 1, |
| "sampling_ratio": 2, |
| "mode": "avg", |
| "coordinate_transformation_mode": "half_pixel", |
| "output_height": 5, |
| "output_width": 5 |
| }, |
| "inputs": { |
| "x": { |
| "dtype": "float32", |
| "shape": [1, 1, 10, 10], |
| "data": { "kind": "values", "values": { "$ref": "#/fixtureArrays/onnx_backend_aligned_false_5x5_input_x" } } |
| }, |
| "rois": { |
| "dtype": "float32", |
| "shape": [3, 4], |
| "data": { "kind": "values", "values": [0.0, 0.0, 9.0, 9.0, 0.0, 5.0, 4.0, 9.0, 5.0, 5.0, 9.0, 9.0] } |
| }, |
| "batch_indices": { "dtype": "uint32", "shape": [3], "data": { "kind": "values", "values": [0, 0, 0] } } |
| }, |
| "outputs": { "y": { "dtype": "float32", "shape": [3, 1, 5, 5], "tolerance": 0.0002 } } |
| }, |
| { |
| "name": "ort_out_of_bounds_degenerate_rois_half_pixel", |
| "provenance": { |
| "source": "onnxruntime/test/providers/cpu/object_detection/roialign_test.cc", |
| "test": "RoiAlignTest.AvgModePositive_half_pixel" |
| }, |
| "attrs": { |
| "spatial_scale": 0.0625, |
| "sampling_ratio": 2, |
| "mode": "avg", |
| "coordinate_transformation_mode": "half_pixel", |
| "output_height": 3, |
| "output_width": 4 |
| }, |
| "inputs": { |
| "x": { |
| "dtype": "float32", |
| "shape": [1, 3, 5, 5], |
| "data": { |
| "kind": "values", |
| "values": { "$ref": "#/fixtureArrays/ort_out_of_bounds_degenerate_rois_half_pixel_input_x" } |
| } |
| }, |
| "rois": { |
| "dtype": "float32", |
| "shape": [5, 4], |
| "data": { |
| "kind": "values", |
| "values": { "$ref": "#/fixtureArrays/ort_out_of_bounds_degenerate_rois_half_pixel_input_rois" } |
| } |
| }, |
| "batch_indices": { "dtype": "uint32", "shape": [5], "data": { "kind": "values", "values": [0, 0, 0, 0, 0] } } |
| }, |
| "outputs": { "y": { "dtype": "float32", "shape": [5, 3, 3, 4], "tolerance": 0.00001 } } |
| }, |
| { |
| "name": "ort_out_of_bounds_degenerate_rois_output_half_pixel", |
| "provenance": { |
| "source": "onnxruntime/test/providers/cpu/object_detection/roialign_test.cc", |
| "test": "RoiAlignTest.AvgModePositive_output_half_pixel" |
| }, |
| "attrs": { |
| "spatial_scale": 0.0625, |
| "sampling_ratio": 2, |
| "mode": "avg", |
| "coordinate_transformation_mode": "output_half_pixel", |
| "output_height": 3, |
| "output_width": 4 |
| }, |
| "inputs": { |
| "x": { |
| "dtype": "float32", |
| "shape": [1, 3, 5, 5], |
| "data": { |
| "kind": "values", |
| "values": { "$ref": "#/fixtureArrays/ort_out_of_bounds_degenerate_rois_half_pixel_input_x" } |
| } |
| }, |
| "rois": { |
| "dtype": "float32", |
| "shape": [5, 4], |
| "data": { |
| "kind": "values", |
| "values": { "$ref": "#/fixtureArrays/ort_out_of_bounds_degenerate_rois_half_pixel_input_rois" } |
| } |
| }, |
| "batch_indices": { "dtype": "uint32", "shape": [5], "data": { "kind": "values", "values": [0, 0, 0, 0, 0] } } |
| }, |
| "outputs": { "y": { "dtype": "float32", "shape": [5, 3, 3, 4], "tolerance": 0.00001 } } |
| }, |
| { |
| "name": "onnx_backend_roialign_aligned_false", |
| "provenance": { |
| "source": "cmake/external/onnx/onnx/backend/test/data/node/test_roialign_aligned_false", |
| "notes": "Representable ONNX int64 batch_indices values are stored in uint32 slots." |
| }, |
| "attrs": { |
| "spatial_scale": 1, |
| "sampling_ratio": 2, |
| "mode": "avg", |
| "coordinate_transformation_mode": "output_half_pixel", |
| "output_height": 5, |
| "output_width": 5 |
| }, |
| "inputs": { |
| "x": { |
| "dtype": "float32", |
| "shape": [1, 1, 10, 10], |
| "data": { "kind": "values", "values": { "$ref": "#/fixtureArrays/onnx_backend_roialign_input_x" } } |
| }, |
| "rois": { |
| "dtype": "float32", |
| "shape": [3, 4], |
| "data": { "kind": "values", "values": [0.0, 0.0, 9.0, 9.0, 0.0, 5.0, 4.0, 9.0, 5.0, 5.0, 9.0, 9.0] } |
| }, |
| "batch_indices": { "dtype": "uint32", "shape": [3], "data": { "kind": "values", "values": [0, 0, 0] } } |
| }, |
| "outputs": { "y": { "dtype": "float32", "shape": [3, 1, 5, 5], "tolerance": 0.0002 } } |
| }, |
| { |
| "name": "onnx_backend_roialign_aligned_true", |
| "provenance": { |
| "source": "cmake/external/onnx/onnx/backend/test/data/node/test_roialign_aligned_true", |
| "notes": "Representable ONNX int64 batch_indices values are stored in uint32 slots." |
| }, |
| "attrs": { |
| "spatial_scale": 1, |
| "sampling_ratio": 2, |
| "mode": "avg", |
| "coordinate_transformation_mode": "half_pixel", |
| "output_height": 5, |
| "output_width": 5 |
| }, |
| "inputs": { |
| "x": { |
| "dtype": "float32", |
| "shape": [1, 1, 10, 10], |
| "data": { "kind": "values", "values": { "$ref": "#/fixtureArrays/onnx_backend_roialign_input_x" } } |
| }, |
| "rois": { |
| "dtype": "float32", |
| "shape": [3, 4], |
| "data": { "kind": "values", "values": [0.0, 0.0, 9.0, 9.0, 0.0, 5.0, 4.0, 9.0, 5.0, 5.0, 9.0, 9.0] } |
| }, |
| "batch_indices": { "dtype": "uint32", "shape": [3], "data": { "kind": "values", "values": [0, 0, 0] } } |
| }, |
| "outputs": { "y": { "dtype": "float32", "shape": [3, 1, 5, 5], "tolerance": 0.0002 } } |
| }, |
| { |
| "name": "onnx_backend_roialign_mode_max", |
| "provenance": { |
| "source": "cmake/external/onnx/onnx/backend/test/data/node/test_roialign_mode_max", |
| "notes": "Representable ONNX int64 batch_indices values are stored in uint32 slots." |
| }, |
| "attrs": { |
| "spatial_scale": 1, |
| "sampling_ratio": 2, |
| "mode": "max", |
| "coordinate_transformation_mode": "output_half_pixel", |
| "output_height": 5, |
| "output_width": 5 |
| }, |
| "inputs": { |
| "x": { |
| "dtype": "float32", |
| "shape": [1, 1, 10, 10], |
| "data": { "kind": "values", "values": { "$ref": "#/fixtureArrays/onnx_backend_roialign_input_x" } } |
| }, |
| "rois": { |
| "dtype": "float32", |
| "shape": [3, 4], |
| "data": { "kind": "values", "values": [0.0, 0.0, 9.0, 9.0, 0.0, 5.0, 4.0, 9.0, 5.0, 5.0, 9.0, 9.0] } |
| }, |
| "batch_indices": { "dtype": "uint32", "shape": [3], "data": { "kind": "values", "values": [0, 0, 0] } } |
| }, |
| "outputs": { "y": { "dtype": "float32", "shape": [3, 1, 5, 5], "tolerance": 0.0002 } } |
| }, |
| { |
| "name": "ort_f16_adaptive_sampling_half_pixel", |
| "provenance": { |
| "source": "onnxruntime/test/providers/cpu/object_detection/roialign_test.cc", |
| "test": "RoiAlignTest.Float16_AdaptiveSampling_Opset16", |
| "notes": "Exercises adaptive per-bin sampling with sampling_ratio=0 and half_pixel coordinates in float16." |
| }, |
| "attrs": { |
| "spatial_scale": 1, |
| "sampling_ratio": 0, |
| "mode": "avg", |
| "coordinate_transformation_mode": "half_pixel", |
| "output_height": 2, |
| "output_width": 2 |
| }, |
| "inputs": { |
| "x": { |
| "dtype": "float16", |
| "shape": [1, 1, 8, 8], |
| "data": { "kind": "values", "values": { "$ref": "#/fixtureArrays/ort_f16_half_pixel_8x8_sampling2_input_x" } } |
| }, |
| "rois": { "dtype": "float16", "shape": [1, 4], "data": { "kind": "values", "values": [0.0, 0.0, 7.0, 7.0] } }, |
| "batch_indices": { "dtype": "uint32", "shape": [1], "data": { "kind": "values", "values": [0] } } |
| }, |
| "outputs": { |
| "y": { |
| "dtype": "float16", |
| "shape": [1, 1, 2, 2], |
| "tolerance": 0.01, |
| "data": { "kind": "values", "values": [11.39062, 14.875, 39.26562, 42.75] } |
| } |
| } |
| }, |
| { |
| "name": "ort_f32_adaptive_sampling_half_pixel", |
| "provenance": { |
| "source": "onnxruntime/test/providers/cpu/object_detection/roialign_test.cc", |
| "test": "RoiAlignTest.Float16_AdaptiveSampling_Opset16", |
| "notes": "Exercises adaptive per-bin sampling with sampling_ratio=0 and half_pixel coordinates in float32." |
| }, |
| "attrs": { |
| "spatial_scale": 1, |
| "sampling_ratio": 0, |
| "mode": "avg", |
| "coordinate_transformation_mode": "half_pixel", |
| "output_height": 2, |
| "output_width": 2 |
| }, |
| "inputs": { |
| "x": { |
| "dtype": "float32", |
| "shape": [1, 1, 8, 8], |
| "data": { "kind": "values", "values": { "$ref": "#/fixtureArrays/ort_f16_half_pixel_8x8_sampling2_input_x" } } |
| }, |
| "rois": { "dtype": "float32", "shape": [1, 4], "data": { "kind": "values", "values": [0.0, 0.0, 7.0, 7.0] } }, |
| "batch_indices": { "dtype": "uint32", "shape": [1], "data": { "kind": "values", "values": [0] } } |
| }, |
| "outputs": { |
| "y": { |
| "dtype": "float32", |
| "shape": [1, 1, 2, 2], |
| "tolerance": 0.00001, |
| "data": { "kind": "values", "values": [11.390625, 14.875, 39.265625, 42.75] } |
| } |
| } |
| }, |
| { |
| "name": "adaptive_avg_output_half_pixel", |
| "provenance": { |
| "source": "onnxruntime/test/providers/cpu/object_detection/roialign_test.cc", |
| "test": "RoiAlignTest.Float16_AdaptiveSampling_Opset16", |
| "notes": "Exercises adaptive average pooling with sampling_ratio=0 and output_half_pixel coordinates." |
| }, |
| "attrs": { |
| "spatial_scale": 1, |
| "sampling_ratio": 0, |
| "mode": "avg", |
| "coordinate_transformation_mode": "output_half_pixel", |
| "output_height": 2, |
| "output_width": 2 |
| }, |
| "inputs": { |
| "x": { |
| "dtype": "float32", |
| "shape": [1, 1, 4, 4], |
| "data": { "kind": "values", "values": { "$ref": "#/fixtureArrays/avg_single_roi_f32_input_x" } } |
| }, |
| "rois": { "dtype": "float32", "shape": [1, 4], "data": { "kind": "values", "values": [0.0, 0.0, 4.0, 4.0] } }, |
| "batch_indices": { "dtype": "uint32", "shape": [1], "data": { "kind": "values", "values": [0] } } |
| }, |
| "outputs": { |
| "y": { |
| "dtype": "float32", |
| "shape": [1, 1, 2, 2], |
| "tolerance": 0.00001, |
| "data": { "kind": "values", "values": [5.0, 6.75, 12.0, 13.75] } |
| } |
| } |
| }, |
| { |
| "name": "ort_max_out_of_bounds_degenerate_rois_output_half_pixel", |
| "provenance": { |
| "source": "onnxruntime/test/providers/cpu/object_detection/roialign_test.cc", |
| "test": "RoiAlignTest.MaxModePositive", |
| "notes": "Opset-10 RoiAlign behavior maps to output_half_pixel in current ONNX attributes." |
| }, |
| "attrs": { |
| "spatial_scale": 0.0625, |
| "sampling_ratio": 2, |
| "mode": "max", |
| "coordinate_transformation_mode": "output_half_pixel", |
| "output_height": 3, |
| "output_width": 4 |
| }, |
| "inputs": { |
| "x": { |
| "dtype": "float32", |
| "shape": [1, 3, 5, 5], |
| "data": { |
| "kind": "values", |
| "values": { "$ref": "#/fixtureArrays/ort_out_of_bounds_degenerate_rois_half_pixel_input_x" } |
| } |
| }, |
| "rois": { |
| "dtype": "float32", |
| "shape": [5, 4], |
| "data": { |
| "kind": "values", |
| "values": { "$ref": "#/fixtureArrays/ort_out_of_bounds_degenerate_rois_half_pixel_input_rois" } |
| } |
| }, |
| "batch_indices": { "dtype": "uint32", "shape": [5], "data": { "kind": "values", "values": [0, 0, 0, 0, 0] } } |
| }, |
| "outputs": { "y": { "dtype": "float32", "shape": [5, 3, 3, 4], "tolerance": 0.00001 } } |
| }, |
| { |
| "name": "empty_input_zero_dim", |
| "attrs": { "spatial_scale": 1, "sampling_ratio": 2, "mode": "avg", "output_height": 4, "output_width": 4 }, |
| "inputs": { |
| "x": { "dtype": "float32", "shape": [0, 1, 4, 4], "data": { "kind": "values", "values": [] } }, |
| "rois": { "dtype": "float32", "shape": [0, 4], "data": { "kind": "values", "values": [] } }, |
| "batch_indices": { "dtype": "uint32", "shape": [0], "data": { "kind": "values", "values": [] } } |
| }, |
| "outputs": { "y": { "dtype": "float32", "shape": [0, 1, 4, 4], "tolerance": 0 } } |
| }, |
| { |
| "name": "adaptive_sampling_zero_constant_input_avg_grid_bounds", |
| "provenance": { |
| "notes": "sampling_ratio=0 selects an adaptive grid with at least one sample per axis. Constant input makes every bilinear sample equal to that constant regardless of the grid, so the inline expected output is exact." |
| }, |
| "attrs": { |
| "spatial_scale": 1, |
| "sampling_ratio": 0, |
| "mode": "avg", |
| "coordinate_transformation_mode": "half_pixel", |
| "output_height": 4, |
| "output_width": 4 |
| }, |
| "inputs": { |
| "x": { "dtype": "float32", "shape": [1, 2, 16, 16], "data": { "kind": "constant", "value": 3.5 } }, |
| "rois": { "dtype": "float32", "shape": [1, 4], "data": { "kind": "values", "values": [1.0, 1.0, 14.0, 14.0] } }, |
| "batch_indices": { "dtype": "uint32", "shape": [1], "data": { "kind": "values", "values": [0] } } |
| }, |
| "outputs": { |
| "y": { |
| "dtype": "float32", |
| "shape": [1, 2, 4, 4], |
| "tolerance": 0.00001, |
| "data": { "kind": "constant", "value": 3.5 } |
| } |
| } |
| }, |
| { |
| "name": "adaptive_sampling_zero_area_roi_grid_clamp_no_divzero", |
| "provenance": { |
| "notes": "For a zero-area ROI with sampling_ratio=0, each zero-sized bin clamps its adaptive grid to one sample, avoiding division by zero. Constant input makes every expected sample exactly 2.0." |
| }, |
| "attrs": { |
| "spatial_scale": 1, |
| "sampling_ratio": 0, |
| "mode": "avg", |
| "coordinate_transformation_mode": "half_pixel", |
| "output_height": 2, |
| "output_width": 2 |
| }, |
| "inputs": { |
| "x": { "dtype": "float32", "shape": [1, 1, 8, 8], "data": { "kind": "constant", "value": 2.0 } }, |
| "rois": { "dtype": "float32", "shape": [1, 4], "data": { "kind": "values", "values": [3.0, 3.0, 3.0, 3.0] } }, |
| "batch_indices": { "dtype": "uint32", "shape": [1], "data": { "kind": "values", "values": [0] } } |
| }, |
| "outputs": { |
| "y": { |
| "dtype": "float32", |
| "shape": [1, 1, 2, 2], |
| "tolerance": 0.00001, |
| "data": { "kind": "constant", "value": 2.0 } |
| } |
| } |
| }, |
| { |
| "name": "dispatch_fold_over_16m_constant_input_exact", |
| "provenance": { |
| "notes": "Y=16,842,752 > 16,777,216 engages the gid.y 2D-fold; constant input 1.0 with fully in-bounds ROIs makes every output exactly 1.0, so fold-arithmetic or last-row-guard bugs corrupt the high rows and fail an exact compare." |
| }, |
| "attrs": { |
| "spatial_scale": 1, |
| "sampling_ratio": 1, |
| "mode": "avg", |
| "coordinate_transformation_mode": "output_half_pixel", |
| "output_height": 256, |
| "output_width": 257 |
| }, |
| "inputs": { |
| "x": { "dtype": "float32", "shape": [1, 4, 32, 32], "data": { "kind": "constant", "value": 1.0 } }, |
| "rois": { "dtype": "float32", "shape": [64, 4], "data": { "kind": "cycle", "values": [0.0, 0.0, 31.0, 31.0] } }, |
| "batch_indices": { "dtype": "uint32", "shape": [64], "data": { "kind": "constant", "value": 0 } } |
| }, |
| "outputs": { |
| "y": { |
| "dtype": "float32", |
| "shape": [64, 4, 256, 257], |
| "tolerance": 0.00001, |
| "data": { "kind": "constant", "value": 1.0 } |
| } |
| } |
| }, |
| { |
| "name": "adaptive_cooperative_constant_64x64_1x1_exact", |
| "provenance": { |
| "notes": "With sampling_ratio=0 on a 64x64 input, one output bin uses an adaptive sample grid of roughly 63x63 points. A constant input of 2.5 makes the cooperative reduction and final divisor exactly observable at tolerance zero." |
| }, |
| "attrs": { |
| "spatial_scale": 1, |
| "sampling_ratio": 0, |
| "mode": "avg", |
| "coordinate_transformation_mode": "half_pixel", |
| "output_height": 1, |
| "output_width": 1 |
| }, |
| "inputs": { |
| "x": { "dtype": "float32", "shape": [1, 1, 64, 64], "data": { "kind": "constant", "value": 2.5 } }, |
| "rois": { "dtype": "float32", "shape": [1, 4], "data": { "kind": "values", "values": [0.0, 0.0, 63.0, 63.0] } }, |
| "batch_indices": { "dtype": "uint32", "shape": [1], "data": { "kind": "values", "values": [0] } } |
| }, |
| "outputs": { |
| "y": { |
| "dtype": "float32", |
| "shape": [1, 1, 1, 1], |
| "tolerance": 0, |
| "data": { "kind": "values", "values": [2.5] } |
| } |
| } |
| }, |
| { |
| "name": "adaptive_cooperative_varying_128x128_2x2", |
| "provenance": { |
| "notes": "A nonconstant input with a 2-by-2 output produces an adaptive grid of about 64-by-64 samples per bin. The input-to-pooled-area ratio of 4,096 exercises dense adaptive sampling." |
| }, |
| "attrs": { |
| "spatial_scale": 1, |
| "sampling_ratio": 0, |
| "mode": "avg", |
| "coordinate_transformation_mode": "half_pixel", |
| "output_height": 2, |
| "output_width": 2 |
| }, |
| "inputs": { |
| "x": { "dtype": "float32", "shape": [1, 1, 128, 128], "data": { "kind": "linspace", "start": 0.0, "end": 4.0 } }, |
| "rois": { |
| "dtype": "float32", |
| "shape": [1, 4], |
| "data": { "kind": "values", "values": [0.0, 0.0, 127.0, 127.0] } |
| }, |
| "batch_indices": { "dtype": "uint32", "shape": [1], "data": { "kind": "values", "values": [0] } } |
| }, |
| "outputs": { "y": { "dtype": "float32", "shape": [1, 1, 2, 2], "tolerance": 0.0001 } } |
| }, |
| { |
| "name": "adaptive_cooperative_max_64x64_1x1", |
| "provenance": { |
| "notes": "Max mode over the adaptive grid exercises cooperative tree reduction. Exact associativity and identical per-sample bilinear arithmetic permit zero tolerance." |
| }, |
| "attrs": { |
| "spatial_scale": 1, |
| "sampling_ratio": 0, |
| "mode": "max", |
| "coordinate_transformation_mode": "half_pixel", |
| "output_height": 1, |
| "output_width": 1 |
| }, |
| "inputs": { |
| "x": { |
| "dtype": "float32", |
| "shape": [1, 1, 64, 64], |
| "data": { "kind": "linspace", "start": -5.0, "end": 100.0 } |
| }, |
| "rois": { "dtype": "float32", "shape": [1, 4], "data": { "kind": "values", "values": [0.0, 0.0, 63.0, 63.0] } }, |
| "batch_indices": { "dtype": "uint32", "shape": [1], "data": { "kind": "values", "values": [0] } } |
| }, |
| "outputs": { "y": { "dtype": "float32", "shape": [1, 1, 1, 1], "tolerance": 0 } } |
| }, |
| { |
| "name": "adaptive_shallow_bins_large_feature_map_scalar", |
| "provenance": { |
| "notes": "A realistic 7x7 pooled ROI on a 128x128 feature map has a 5x5 adaptive grid per bin and remains on the one-thread-per-output path. Constant input makes every expected output exactly 3.0." |
| }, |
| "attrs": { |
| "spatial_scale": 1, |
| "sampling_ratio": 0, |
| "mode": "avg", |
| "coordinate_transformation_mode": "half_pixel", |
| "output_height": 7, |
| "output_width": 7 |
| }, |
| "inputs": { |
| "x": { "dtype": "float32", "shape": [1, 1, 128, 128], "data": { "kind": "constant", "value": 3.0 } }, |
| "rois": { |
| "dtype": "float32", |
| "shape": [1, 4], |
| "data": { "kind": "values", "values": [16.0, 16.0, 48.0, 48.0] } |
| }, |
| "batch_indices": { "dtype": "uint32", "shape": [1], "data": { "kind": "values", "values": [0] } } |
| }, |
| "outputs": { |
| "y": { |
| "dtype": "float32", |
| "shape": [1, 1, 7, 7], |
| "tolerance": 0.00001, |
| "data": { "kind": "constant", "value": 3.0 } |
| } |
| } |
| }, |
| { |
| "name": "adaptive_split_varying_257x257_1x1_channel_tail", |
| "provenance": { |
| "notes": "A non-divisible 256-by-255 adaptive grid over three output channels creates many empty split partitions and verifies partial normalization and combine indexing." |
| }, |
| "attrs": { |
| "spatial_scale": 1, |
| "sampling_ratio": 0, |
| "mode": "avg", |
| "coordinate_transformation_mode": "half_pixel", |
| "output_height": 1, |
| "output_width": 1 |
| }, |
| "inputs": { |
| "x": { |
| "dtype": "float32", |
| "shape": [1, 3, 257, 257], |
| "data": { "kind": "linspace", "start": -2.0, "end": 5.0 } |
| }, |
| "rois": { |
| "dtype": "float32", |
| "shape": [1, 4], |
| "data": { "kind": "values", "values": [0.25, 1.5, 256.0, 256.0] } |
| }, |
| "batch_indices": { "dtype": "uint32", "shape": [1], "data": { "kind": "values", "values": [0] } } |
| }, |
| "outputs": { "y": { "dtype": "float32", "shape": [1, 3, 1, 1], "tolerance": 0.0002 } } |
| }, |
| { |
| "name": "adaptive_split_max_256x256_1x1", |
| "provenance": { |
| "notes": "A 256x256 adaptive grid exercises the max-mode split combine. The small tolerance covers float32 contraction and rounding in bilinear sample evaluation without hiding an incorrect maximum." |
| }, |
| "attrs": { |
| "spatial_scale": 1, |
| "sampling_ratio": 0, |
| "mode": "max", |
| "coordinate_transformation_mode": "half_pixel", |
| "output_height": 1, |
| "output_width": 1 |
| }, |
| "inputs": { |
| "x": { |
| "dtype": "float32", |
| "shape": [1, 1, 256, 256], |
| "data": { "kind": "linspace", "start": -3.0, "end": 9.0 } |
| }, |
| "rois": { |
| "dtype": "float32", |
| "shape": [1, 4], |
| "data": { "kind": "values", "values": [0.0, 0.0, 255.0, 255.0] } |
| }, |
| "batch_indices": { "dtype": "uint32", "shape": [1], "data": { "kind": "values", "values": [0] } } |
| }, |
| "outputs": { "y": { "dtype": "float32", "shape": [1, 1, 1, 1], "tolerance": 0.0005 } } |
| }, |
| { |
| "name": "adaptive_cooperative_constant_64x64_1x1_exact_f16", |
| "provenance": { |
| "notes": "A float16 cooperative adaptive-sampling case uses constant 2.5 input, so every bilinear sample and the reduced average are exactly 2.5 regardless of reduction order." |
| }, |
| "attrs": { |
| "spatial_scale": 1, |
| "sampling_ratio": 0, |
| "mode": "avg", |
| "coordinate_transformation_mode": "half_pixel", |
| "output_height": 1, |
| "output_width": 1 |
| }, |
| "inputs": { |
| "x": { "dtype": "float16", "shape": [1, 1, 64, 64], "data": { "kind": "constant", "value": 2.5 } }, |
| "rois": { "dtype": "float16", "shape": [1, 4], "data": { "kind": "values", "values": [0.0, 0.0, 63.0, 63.0] } }, |
| "batch_indices": { "dtype": "uint32", "shape": [1], "data": { "kind": "values", "values": [0] } } |
| }, |
| "outputs": { |
| "y": { |
| "dtype": "float16", |
| "shape": [1, 1, 1, 1], |
| "tolerance": 0, |
| "data": { "kind": "values", "values": [2.5] } |
| } |
| } |
| }, |
| { |
| "name": "adaptive_split_constant_256x256_1x1_exact_f16", |
| "provenance": { |
| "notes": "A float16 256-by-256 input pooled to 1-by-1 reaches the 65,536 input-to-output area ratio for split adaptive sampling. Constant input makes the average exact, verifying split partials and their combine at zero tolerance." |
| }, |
| "attrs": { |
| "spatial_scale": 1, |
| "sampling_ratio": 0, |
| "mode": "avg", |
| "coordinate_transformation_mode": "half_pixel", |
| "output_height": 1, |
| "output_width": 1 |
| }, |
| "inputs": { |
| "x": { "dtype": "float16", "shape": [1, 1, 256, 256], "data": { "kind": "constant", "value": 2.5 } }, |
| "rois": { |
| "dtype": "float16", |
| "shape": [1, 4], |
| "data": { "kind": "values", "values": [0.0, 0.0, 255.0, 255.0] } |
| }, |
| "batch_indices": { "dtype": "uint32", "shape": [1], "data": { "kind": "values", "values": [0] } } |
| }, |
| "outputs": { |
| "y": { |
| "dtype": "float16", |
| "shape": [1, 1, 1, 1], |
| "tolerance": 0, |
| "data": { "kind": "values", "values": [2.5] } |
| } |
| } |
| }, |
| { |
| "name": "max_all_negative_infinity_samples", |
| "provenance": { |
| "notes": "Every bilinear weight at this single centre sample is strictly positive, so the pooled value is exactly -Infinity with no 0 * Infinity ambiguity. The max reductions used a finite -FLT_MAX identity that a sampled -Infinity could not replace, and reported the sentinel instead." |
| }, |
| "attrs": { |
| "mode": "max", |
| "sampling_ratio": 1, |
| "coordinate_transformation_mode": "output_half_pixel", |
| "output_height": 1, |
| "output_width": 1 |
| }, |
| "inputs": { |
| "x": { |
| "dtype": "float32", |
| "shape": [1, 1, 2, 2], |
| "data": { "kind": "values", "values": ["-Infinity", "-Infinity", "-Infinity", "-Infinity"] } |
| }, |
| "rois": { "dtype": "float32", "shape": [1, 4], "data": { "kind": "values", "values": [0.0, 0.0, 1.0, 1.0] } }, |
| "batch_indices": { "dtype": "uint32", "shape": [1], "data": { "kind": "values", "values": [0] } } |
| }, |
| "outputs": { "y": { "dtype": "float32", "shape": [1, 1, 1, 1], "tolerance": 0 } } |
| } |
| ] |
| } |
|
|
