| { |
| "op": "ai.onnx.NonZero", |
| "cases": [ |
| { |
| "name": "int16_scalar_boundaries", |
| "inputs": { |
| "x": { "dtype": "int16", "shape": [4], "data": { "kind": "values", "values": [-32768, 0, 32767, -1] } } |
| }, |
| "outputs": { |
| "y": { "dtype": "uint32", "shape": [1, 3], "tolerance": 0, "data": { "kind": "values", "values": [0, 2, 3] } } |
| } |
| }, |
| { |
| "name": "rank2_exact_capacity_f32", |
| "inputs": { |
| "x": { |
| "dtype": "float32", |
| "shape": [2, 3], |
| "data": { "kind": "values", "values": [0.0, 1.0, 0.0, -2.0, 3.0, 0.0] } |
| } |
| }, |
| "outputs": { "y": { "dtype": "uint32", "shape": [2, 3] } } |
| }, |
| { |
| "name": "f32_subnormal_nonzero_coordinates", |
| "provenance": { |
| "source": "onnxruntime/test/providers/cpu/tensor/nonzero_op_test.cc", |
| "test": "NonZeroOpTest.BasicNumeric", |
| "notes": "Subnormal finite values are nonzero under ONNX equality semantics; flushing them to zero changes the emitted coordinates." |
| }, |
| "inputs": { |
| "x": { |
| "dtype": "float32", |
| "shape": [2, 3], |
| "data": { "kind": "values", "values": [0.0, 1e-40, -1e-40, 0.0, 1e-39, 0.0] } |
| } |
| }, |
| "outputs": { |
| "y": { |
| "dtype": "uint32", |
| "shape": [2, 3], |
| "tolerance": 0, |
| "data": { "kind": "values", "values": [0, 0, 1, 1, 2, 1] } |
| } |
| } |
| }, |
| { |
| "name": "f32_scalar_subnormal_nonzero_coordinates", |
| "provenance": { |
| "source": "onnxruntime/test/providers/cpu/tensor/nonzero_op_test.cc", |
| "test": "NonZeroOpTest.Scalar", |
| "notes": "A finite subnormal rank-0 input is still nonzero. ONNX represents its one hit with shape [0, 1], containing no coordinate values." |
| }, |
| "inputs": { "x": { "dtype": "float32", "shape": [], "data": { "kind": "values", "values": [1e-40] } } }, |
| "outputs": { |
| "y": { "dtype": "uint32", "shape": [0, 1], "tolerance": 0, "data": { "kind": "values", "values": [] } } |
| } |
| }, |
| { |
| "name": "f32_parallel_subnormal_nonzero_coordinates", |
| "provenance": { |
| "source": "onnxruntime/test/providers/cpu/tensor/nonzero_op_test.cc", |
| "test": "NonZeroOpTest.BasicNumeric", |
| "notes": "Large enough to select the parallel-scan path; every even column is a finite subnormal and must produce an output coordinate." |
| }, |
| "inputs": { "x": { "dtype": "float32", "shape": [1, 1026], "data": { "kind": "cycle", "values": [1e-40, 0.0] } } }, |
| "outputs": { "y": { "dtype": "uint32", "shape": [2, 513], "tolerance": 0 } } |
| }, |
| { |
| "name": "rank3_exact_capacity_uint8", |
| "inputs": { |
| "x": { "dtype": "uint8", "shape": [2, 2, 2], "data": { "kind": "values", "values": [1, 0, 2, 3, 0, 4, 5, 6] } } |
| }, |
| "outputs": { "y": { "dtype": "uint32", "shape": [3, 6] } } |
| }, |
| { |
| "name": "rank1_no_hits_int8", |
| "inputs": { "x": { "dtype": "int8", "shape": [4], "data": { "kind": "values", "values": [0, 0, 0, 0] } } }, |
| "outputs": { "y": { "dtype": "uint32", "shape": [1, 0] } } |
| }, |
| { |
| "name": "rank1_nonzero_int8", |
| "inputs": { "x": { "dtype": "int8", "shape": [4], "data": { "kind": "values", "values": [-2, 0, 7, 0] } } }, |
| "outputs": { |
| "y": { "dtype": "uint32", "shape": [1, 2], "tolerance": 0, "data": { "kind": "values", "values": [0, 2] } } |
| }, |
| "provenance": { |
| "notes": "Synthetic compact non-empty int8 witness for the standard numeric NonZero route; the existing int8 zero-hit case does not dispatch the dtype-specific shader path." |
| } |
| }, |
| { |
| "name": "rank4_exact_capacity_f32", |
| "inputs": { |
| "x": { |
| "dtype": "float32", |
| "shape": [1, 2, 2, 3], |
| "data": { "kind": "values", "values": [0.0, 1.0, 2.0, 0.0, 0.0, -3.0, 4.0, 0.0, 5.0, 6.0, 0.0, 7.0] } |
| } |
| }, |
| "outputs": { "y": { "dtype": "uint32", "shape": [4, 7] } } |
| }, |
| { |
| "name": "ort_basic_numeric_rank3_exact_capacity_int32", |
| "provenance": { |
| "source": "onnxruntime/test/providers/cpu/tensor/nonzero_op_test.cc", |
| "test": "NonZeroOpTest.BasicNumeric", |
| "notes": "The runtime represents ONNX int64 coordinate metadata as uint32." |
| }, |
| "inputs": { |
| "x": { "dtype": "int32", "shape": [1, 2, 3], "data": { "kind": "values", "values": [0, 1, 2, 0, 3, 4] } } |
| }, |
| "outputs": { "y": { "dtype": "uint32", "shape": [3, 4], "tolerance": 0 } } |
| }, |
| { |
| "name": "ort_basic_numeric_rank3_exact_capacity_float", |
| "provenance": { |
| "source": "onnxruntime/test/providers/cpu/tensor/nonzero_op_test.cc", |
| "test": "NonZeroOpTest.BasicNumeric", |
| "notes": "Float variant from ORT’s templated BasicNumeric case; the runtime represents ONNX int64 coordinate metadata as uint32." |
| }, |
| "inputs": { |
| "x": { |
| "dtype": "float32", |
| "shape": [1, 2, 3], |
| "data": { "kind": "values", "values": [0.0, 1.0, 2.0, 0.0, 3.0, 4.0] } |
| } |
| }, |
| "outputs": { "y": { "dtype": "uint32", "shape": [3, 4], "tolerance": 0 } } |
| }, |
| { |
| "name": "ort_three_dims_rank3_exact_capacity_int32", |
| "provenance": { |
| "source": "onnxruntime/test/providers/cpu/tensor/nonzero_op_test.cc", |
| "test": "NonZeroOpTest.ThreeDims", |
| "notes": "Integer-width adapted to int32; the runtime represents ONNX int64 coordinate metadata as uint32." |
| }, |
| "inputs": { |
| "x": { "dtype": "int32", "shape": [2, 2, 2], "data": { "kind": "values", "values": [0, 1, 1, 0, 1, 0, 1, 0] } } |
| }, |
| "outputs": { "y": { "dtype": "uint32", "shape": [3, 4], "tolerance": 0 } } |
| }, |
| { |
| "name": "ort_empty_input_rank3_zero_hits", |
| "provenance": { |
| "source": "onnxruntime/test/providers/cpu/tensor/nonzero_op_test.cc", |
| "test": "NonZeroOpTest.EmptyInput", |
| "notes": "Empty-input coverage with the exact ONNX output shape [rank(X), 0]." |
| }, |
| "inputs": { "x": { "dtype": "int32", "shape": [1, 0, 2], "data": { "kind": "values", "values": [] } } }, |
| "outputs": { "y": { "dtype": "uint32", "shape": [3, 0], "tolerance": 0 } } |
| }, |
| { |
| "name": "ort_basic_bool_adapted_int32", |
| "provenance": { |
| "source": "onnxruntime/test/providers/cpu/tensor/nonzero_op_test.cc", |
| "test": "NonZeroOpTest.BasicBool", |
| "notes": "Integer twin of the boolean pattern: a [2,3] int32 0/1 tensor pins row-major coordinates for its two nonzero elements and the uint32 metadata output." |
| }, |
| "inputs": { |
| "x": { "dtype": "int32", "shape": [2, 3], "data": { "kind": "values", "values": [1, 0, 0, 0, 0, 1] } } |
| }, |
| "outputs": { "y": { "dtype": "uint32", "shape": [2, 2], "tolerance": 0 } } |
| }, |
| { |
| "name": "ort_basic_bool_exact", |
| "provenance": { |
| "source": "onnxruntime/test/providers/cpu/tensor/nonzero_op_test.cc", |
| "test": "NonZeroOpTest.BasicBool" |
| }, |
| "inputs": { |
| "x": { "dtype": "bool", "shape": [2, 3], "data": { "kind": "values", "values": [1, 0, 0, 0, 0, 1] } } |
| }, |
| "outputs": { |
| "y": { |
| "dtype": "uint32", |
| "shape": [2, 2], |
| "tolerance": 0, |
| "data": { "kind": "values", "values": [0, 1, 0, 2] } |
| } |
| } |
| }, |
| { |
| "name": "ort_scalar_zero_int32", |
| "provenance": { |
| "source": "onnxruntime/test/providers/cpu/tensor/nonzero_op_test.cc", |
| "test": "NonZeroOpTest.Scalar", |
| "notes": "Pins the ONNX scalar-zero output shape [0, 0]; the runtime represents ONNX int64 coordinate metadata as uint32." |
| }, |
| "inputs": { "x": { "dtype": "int32", "shape": [], "data": { "kind": "values", "values": [0] } } }, |
| "outputs": { "y": { "dtype": "uint32", "shape": [0, 0], "tolerance": 0 } } |
| }, |
| { |
| "name": "ort_scalar_nonzero_int32", |
| "provenance": { |
| "source": "onnxruntime/test/providers/cpu/tensor/nonzero_op_test.cc", |
| "test": "NonZeroOpTest.Scalar", |
| "notes": "Pins the ONNX scalar-nonzero output shape [0, 1]; the runtime represents ONNX int64 coordinate metadata as uint32." |
| }, |
| "inputs": { "x": { "dtype": "int32", "shape": [], "data": { "kind": "values", "values": [1] } } }, |
| "outputs": { "y": { "dtype": "uint32", "shape": [0, 1], "tolerance": 0 } } |
| }, |
| { |
| "name": "bool_scalar_nonzero", |
| "provenance": { |
| "source": "onnxruntime/test/providers/cpu/tensor/nonzero_op_test.cc", |
| "test": "NonZeroOpTest.Scalar", |
| "notes": "Scalar NonZero shape convention from ORT's scalar case, using ONNX-valid bool input coverage from NonZeroOpTest.BasicBool." |
| }, |
| "inputs": { "x": { "dtype": "bool", "shape": [], "data": { "kind": "values", "values": [1] } } }, |
| "outputs": { |
| "y": { "dtype": "uint32", "shape": [0, 1], "tolerance": 0, "data": { "kind": "values", "values": [] } } |
| } |
| }, |
| { |
| "name": "onnx_backend_nonzero_example", |
| "inputs": { "x": { "dtype": "int32", "shape": [2, 2], "data": { "kind": "values", "values": [1, 0, 1, 1] } } }, |
| "outputs": { "y": { "dtype": "uint32", "shape": [2, 3] } }, |
| "provenance": { |
| "source": "cmake/external/onnx/onnx/backend/test/data/node/test_nonzero_example", |
| "notes": "Official ONNX example with its exact dynamic output shape. The runtime represents ONNX int64 coordinate metadata as uint32." |
| } |
| }, |
| { |
| "name": "float32_signed_zero_nan_and_infinity", |
| "provenance": { |
| "source": "onnxruntime/test/providers/cpu/tensor/nonzero_op_test.cc", |
| "test": "NonZeroOpTest.BasicNumeric", |
| "notes": "Additional numeric edge: +0 and -0 are zero, while NaN and infinities compare nonzero." |
| }, |
| "inputs": { |
| "x": { |
| "dtype": "float32", |
| "shape": [2, 3], |
| "data": { "kind": "values", "values": [0.0, 0.0, "NaN", "Infinity", "-Infinity", 1.0] } |
| } |
| }, |
| "outputs": { |
| "y": { |
| "dtype": "uint32", |
| "shape": [2, 4], |
| "tolerance": 0, |
| "data": { "kind": "values", "values": [0, 1, 1, 1, 2, 0, 1, 2] } |
| } |
| } |
| }, |
| { |
| "name": "float16_signed_zero_nan_exact_capacity", |
| "provenance": { |
| "source": "onnxruntime/test/providers/cpu/tensor/nonzero_op_test.cc", |
| "test": "NonZeroOpTest.BasicNumeric", |
| "notes": "Additional numeric edge: -0 is zero, while NaN and the finite nonzero value produce the two exact output coordinates." |
| }, |
| "inputs": { |
| "x": { "dtype": "float16", "shape": [4], "data": { "kind": "values", "values": [0.0, "NaN", 0.0, -2.0] } } |
| }, |
| "outputs": { |
| "y": { "dtype": "uint32", "shape": [1, 2], "tolerance": 0, "data": { "kind": "values", "values": [1, 3] } } |
| } |
| }, |
| { |
| "name": "parallel_rank2_exact_capacity_f32", |
| "provenance": { |
| "notes": "65536-element rank-2 input with 25% nonzeros at exact capacity; exercises the parallel_scan 3-pass coordinate-emission variant." |
| }, |
| "inputs": { |
| "x": { |
| "dtype": "float32", |
| "shape": [256, 256], |
| "data": { "kind": "cycle", "values": [0.0, 0.0, 2.5, 0.0, -1.25, 0.0, 0.0, 0.0] } |
| } |
| }, |
| "outputs": { "y": { "dtype": "uint32", "shape": [2, 16384], "tolerance": 0 } } |
| }, |
| { |
| "name": "parallel_rank1_f32_subnormal_nonzero", |
| "provenance": { |
| "source": "onnxruntime/test/providers/cpu/tensor/nonzero_op_test.cc", |
| "test": "NonZeroOpTest.BasicNumeric", |
| "notes": "Parallel-scan companion to the serial subnormal case: finite subnormal float32 values are nonzero and must survive the shared predicate scan." |
| }, |
| "inputs": { |
| "x": { "dtype": "float32", "shape": [4096], "data": { "kind": "cycle", "values": [0.0, 1e-40, 0.0, -1e-40] } } |
| }, |
| "outputs": { "y": { "dtype": "uint32", "shape": [1, 2048], "tolerance": 0 } } |
| }, |
| { |
| "name": "parallel_rank1_exact_capacity_int32", |
| "provenance": { |
| "notes": "Exactly 10000 nonzeros; the parallel scatter must emit every coordinate into the caller-provided exact output shape." |
| }, |
| "inputs": { "x": { "dtype": "int32", "shape": [50000], "data": { "kind": "cycle", "values": [0, 3, 0, 0, 0] } } }, |
| "outputs": { "y": { "dtype": "uint32", "shape": [1, 10000], "tolerance": 0 } } |
| }, |
| { |
| "name": "parallel_rank3_u32", |
| "provenance": { |
| "notes": "Rank-3 uint32 input; render-time stride unrolling must emit three coordinate rows at the scanned columns." |
| }, |
| "inputs": { |
| "x": { "dtype": "uint32", "shape": [32, 60, 8], "data": { "kind": "cycle", "values": [7, 0, 0, 0, 1, 0] } } |
| }, |
| "outputs": { "y": { "dtype": "uint32", "shape": [3, 5120], "tolerance": 0 } } |
| }, |
| { |
| "name": "parallel_rank1_f16", |
| "provenance": { |
| "notes": "f16 input through the parallel predicate scan (f32-widened zero test, same as the serial kernel)." |
| }, |
| "inputs": { |
| "x": { "dtype": "float16", "shape": [4096], "data": { "kind": "cycle", "values": [0.0, 1.5, 0.0, 0.0] } } |
| }, |
| "outputs": { "y": { "dtype": "uint32", "shape": [1, 1024], "tolerance": 0 } } |
| }, |
| { |
| "name": "f16_subnormal_nonzero_coordinates", |
| "provenance": { |
| "source": "onnxruntime/test/providers/cpu/tensor/nonzero_op_test.cc", |
| "test": "NonZeroOpTest.BasicNumeric", |
| "notes": "The f16 predicate widens values to f32 before comparing with zero. The three finite f16 subnormals must therefore produce coordinates (0,1), (0,2), and (1,1), proving that widening preserves their nonzero values." |
| }, |
| "inputs": { |
| "x": { |
| "dtype": "float16", |
| "shape": [2, 3], |
| "data": { "kind": "values", "values": [0.0, 6e-8, -6e-8, 0.0, 0.00006, 0.0] } |
| } |
| }, |
| "outputs": { |
| "y": { |
| "dtype": "uint32", |
| "shape": [2, 3], |
| "tolerance": 0, |
| "data": { "kind": "values", "values": [0, 0, 1, 1, 2, 1] } |
| } |
| } |
| }, |
| { |
| "name": "parallel_f16_subnormal_nonzero_coordinates", |
| "provenance": { |
| "source": "onnxruntime/test/providers/cpu/tensor/nonzero_op_test.cc", |
| "test": "NonZeroOpTest.BasicNumeric", |
| "notes": "Parallel companion to the serial f16-subnormal case. Every even element rounds to the finite f16 subnormal 0x0001 and must survive the shared widened-f32 predicate, producing exactly 1024 coordinate columns." |
| }, |
| "inputs": { "x": { "dtype": "float16", "shape": [2048], "data": { "kind": "cycle", "values": [6e-8, 0.0] } } }, |
| "outputs": { "y": { "dtype": "uint32", "shape": [1, 1024], "tolerance": 0 } } |
| }, |
| { |
| "name": "parallel_rank1_bool_mask", |
| "provenance": { |
| "source": "onnxruntime/test/providers/cpu/tensor/nonzero_op_test.cc", |
| "test": "NonZeroOpTest.BasicBool", |
| "notes": "Clean coverage: a sequence-scale boolean mask (the canonical real-model NonZero producer) through the parallel_scan path (numel>1024). bool maps to the u32 scalar with predicate `src[i] != 0u`; only the serial variant carried bool coverage before. ORT CPU NonZero(bool) and the TS reference agree on 1024 hits at the even columns." |
| }, |
| "inputs": { "x": { "dtype": "bool", "shape": [2048], "data": { "kind": "cycle", "values": [1, 0] } } }, |
| "outputs": { "y": { "dtype": "uint32", "shape": [1, 1024], "tolerance": 0 } } |
| }, |
| { |
| "name": "parallel_rank2_uint8", |
| "provenance": { |
| "source": "onnxruntime/test/providers/cpu/tensor/nonzero_op_test.cc", |
| "test": "NonZeroOpTest.BasicNumeric", |
| "notes": "Clean coverage: rank-2 uint8 input through the parallel_scan path (numel>1024); uint8 maps to the u32 scalar with predicate `src[i] != 0u`. uint8 NonZero was only exercised by the tiny serial case before. ORT CPU NonZero(uint8) and the TS reference agree (every third element nonzero)." |
| }, |
| "inputs": { "x": { "dtype": "uint8", "shape": [30, 50], "data": { "kind": "cycle", "values": [7, 0, 0] } } }, |
| "outputs": { "y": { "dtype": "uint32", "shape": [2, 500], "tolerance": 0 } } |
| }, |
| { |
| "name": "parallel_f32_signed_zero_nan_infinity", |
| "provenance": { |
| "source": "onnxruntime/test/providers/cpu/tensor/nonzero_op_test.cc", |
| "test": "NonZeroOpTest.BasicNumeric", |
| "notes": "Parallel companion to float32_signed_zero_nan_and_infinity. The flag scan bit-tests `(bitcast<u32>(src[i]) & 0x7fffffffu) != 0u`, so +0/-0 are zero while NaN and +/-Inf are nonzero; the six-value cycle produces exactly 1024 coordinate columns." |
| }, |
| "inputs": { |
| "x": { |
| "dtype": "float32", |
| "shape": [1536], |
| "data": { "kind": "cycle", "values": [0.0, 0.0, "NaN", "Infinity", "-Infinity", 1.0] } |
| } |
| }, |
| "outputs": { "y": { "dtype": "uint32", "shape": [1, 1024], "tolerance": 0 } } |
| }, |
| { |
| "name": "parallel_rank6_max_rank_stride_unroll", |
| "provenance": { |
| "notes": "Six-dimensional int32 input through the parallel path. This exercises six compile-time coordinate strides and emits six coordinate rows; the [3,0,0,0] cycle produces 512 deterministic columns in row-major order." |
| }, |
| "inputs": { |
| "x": { "dtype": "int32", "shape": [2, 2, 2, 2, 2, 64], "data": { "kind": "cycle", "values": [3, 0, 0, 0] } } |
| }, |
| "outputs": { "y": { "dtype": "uint32", "shape": [6, 512], "tolerance": 0 } } |
| }, |
| { |
| "name": "parallel_rank4_vec4_tail_scan_scatter", |
| "provenance": { |
| "notes": "Compact route lock for the rank-4 activation-map benchmark. 1536 elements cross one 1024-element vectorized block boundary; the 50% mask validates stable flattened order, rank-4 coordinate decomposition, and the partial second block." |
| }, |
| "inputs": { "x": { "dtype": "uint32", "shape": [2, 3, 8, 32], "data": { "kind": "cycle", "values": [0, 7] } } }, |
| "outputs": { "y": { "dtype": "uint32", "shape": [4, 768], "tolerance": 0 } } |
| }, |
| { |
| "name": "parallel_rank1_exact_crossover_1024", |
| "provenance": { |
| "notes": "Pins the first profitable parallel-scan size. Exactly 1024 elements form one complete vectorized scan block; the alternating mask validates stable coordinates and an exact 512-column output at the serial/parallel selector boundary." |
| }, |
| "inputs": { "x": { "dtype": "float32", "shape": [1024], "data": { "kind": "cycle", "values": [0.0, 3.0] } } }, |
| "outputs": { "y": { "dtype": "uint32", "shape": [1, 512], "tolerance": 0 } } |
| }, |
| { |
| "name": "rank7_exact_capacity_coordinates", |
| "inputs": { |
| "x": { |
| "dtype": "float32", |
| "shape": [1, 2, 1, 2, 1, 2, 3], |
| "data": { "kind": "cycle", "values": [0.0, 1.0, 0.0, -2.0] } |
| } |
| }, |
| "outputs": { "y": { "dtype": "uint32", "shape": [7, 12], "tolerance": 0 } } |
| }, |
| { |
| "name": "parallel_sparse_exact_capacity_2048", |
| "provenance": { |
| "notes": "Sparse parallel-route fixture with 512 deterministic hits and an exact [1, 512] output shape." |
| }, |
| "inputs": { |
| "x": { "dtype": "float32", "shape": [2048], "data": { "kind": "cycle", "values": [0.0, 3.0, 0.0, 0.0] } } |
| }, |
| "outputs": { "y": { "dtype": "uint32", "shape": [1, 512], "tolerance": 0 } } |
| }, |
| { |
| "name": "rank8_exact_capacity_coordinates", |
| "inputs": { |
| "x": { |
| "dtype": "float32", |
| "shape": [1, 2, 1, 2, 1, 2, 2, 3], |
| "data": { "kind": "cycle", "values": [0.0, 1.0, 0.0, -2.0] } |
| } |
| }, |
| "outputs": { "y": { "dtype": "uint32", "shape": [8, 24], "tolerance": 0 } } |
| }, |
| { |
| "name": "parallel_rank4_wide_items_scan", |
| "provenance": { |
| "notes": "Large enough (524288 elements, 512 scan blocks) to take the wide items-per-thread branch, which for rank >= 4 resolves to 4. Every other parallel fixture is small enough to fall to the narrow 16-item branch, so without this the 4-item scan and scatter were exercised only by benchmarks, which do not check output." |
| }, |
| "inputs": { |
| "x": { |
| "dtype": "uint32", |
| "shape": [4, 8, 128, 128], |
| "data": { "kind": "cycle", "values": [0, 0, 0, 0, 0, 0, 0, 7] } |
| } |
| }, |
| "outputs": { "y": { "dtype": "uint32", "shape": [4, 65536], "tolerance": 0 } } |
| }, |
| { |
| "name": "parallel_rank3_wide_items_scan", |
| "provenance": { |
| "notes": "Rank-3 twin of parallel_rank4_wide_items_scan: 1048576 elements is 512 scan blocks at the rank-3 wide value of 8 items per thread, the other items-per-thread value no smaller fixture reaches." |
| }, |
| "inputs": { |
| "x": { |
| "dtype": "uint32", |
| "shape": [8, 128, 1024], |
| "data": { "kind": "cycle", "values": [0, 0, 0, 0, 0, 0, 0, 7] } |
| } |
| }, |
| "outputs": { "y": { "dtype": "uint32", "shape": [3, 131072], "tolerance": 0 } } |
| } |
| ] |
| } |
|
|
