|
| 1 | +/* |
| 2 | + * Copyright (c) 2025 Arm Limited. |
| 3 | + * |
| 4 | + * SPDX-License-Identifier: MIT |
| 5 | + * |
| 6 | + * Permission is hereby granted, free of charge, to any person obtaining a copy |
| 7 | + * of this software and associated documentation files (the "Software"), to |
| 8 | + * deal in the Software without restriction, including without limitation the |
| 9 | + * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or |
| 10 | + * sell copies of the Software, and to permit persons to whom the Software is |
| 11 | + * furnished to do so, subject to the following conditions: |
| 12 | + * |
| 13 | + * The above copyright notice and this permission notice shall be included in all |
| 14 | + * copies or substantial portions of the Software. |
| 15 | + * |
| 16 | + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR |
| 17 | + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, |
| 18 | + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE |
| 19 | + * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER |
| 20 | + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, |
| 21 | + * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE |
| 22 | + * SOFTWARE. |
| 23 | + */ |
| 24 | +#include "arm_compute/core/TensorFormat.h" |
| 25 | +#include "arm_compute/core/Types.h" |
| 26 | +#include "arm_compute/runtime/Tensor.h" |
| 27 | +#include "arm_compute/runtime/COOTensor.h" |
| 28 | +#include "tests/framework/Asserts.h" |
| 29 | +#include "tests/framework/Macros.h" |
| 30 | +#include "tests/framework/datasets/Datasets.h" |
| 31 | +#include "tests/validation/Validation.h" |
| 32 | +#include "tests/validation/Helpers.h" |
| 33 | + |
| 34 | +#include "tests/NEON/Accessor.h" |
| 35 | +#include "tests/NEON/Helper.h" |
| 36 | + |
| 37 | +#include <vector> |
| 38 | + |
| 39 | +namespace arm_compute |
| 40 | +{ |
| 41 | +namespace |
| 42 | +{ |
| 43 | +bool are_values_equal(const uint8_t *a, const uint8_t *b, DataType dt, size_t element_size) |
| 44 | +{ |
| 45 | + if(dt == DataType::F32) |
| 46 | + { |
| 47 | + float va = *reinterpret_cast<const float *>(a); |
| 48 | + float vb = *reinterpret_cast<const float *>(b); |
| 49 | + if(std::fabs(va - vb) > 0e-5f) |
| 50 | + { |
| 51 | + return false; |
| 52 | + } |
| 53 | + } else |
| 54 | + { |
| 55 | + if(std::memcmp(a, b, element_size) != 0) |
| 56 | + { |
| 57 | + return false; |
| 58 | + } |
| 59 | + } |
| 60 | + |
| 61 | + return true; |
| 62 | +} |
| 63 | + |
| 64 | +bool tensors_are_equal(const test::Accessor &a, const test::Accessor &b) |
| 65 | +{ |
| 66 | + if(a.shape() != b.shape() || a.data_type() != b.data_type()) |
| 67 | + return false; |
| 68 | + |
| 69 | + const size_t element_size = a.element_size(); |
| 70 | + Window window; |
| 71 | + window.use_tensor_dimensions(a.shape()); |
| 72 | + |
| 73 | + bool equal = true; |
| 74 | + |
| 75 | + execute_window_loop(window, [&](const Coordinates &id) |
| 76 | + { |
| 77 | + const uint8_t *a_value = static_cast<const uint8_t *>(a(id)); |
| 78 | + const uint8_t *b_value = static_cast<const uint8_t *>(b(id)); |
| 79 | + |
| 80 | + equal = are_values_equal(a_value, b_value, a.data_type(), element_size); |
| 81 | + }); |
| 82 | + |
| 83 | + return equal; |
| 84 | +} |
| 85 | +} // namespace |
| 86 | + |
| 87 | +namespace test |
| 88 | +{ |
| 89 | +namespace validation |
| 90 | +{ |
| 91 | +TEST_SUITE(UNIT) |
| 92 | +TEST_SUITE(SparseTensor) |
| 93 | + |
| 94 | +// clang-format off |
| 95 | +/** Validates TensorInfo Autopadding */ |
| 96 | +DATA_TEST_CASE(ConvertCOOTensorToDense, framework::DatasetMode::ALL, combine( |
| 97 | + framework::dataset::make("TensorShape", { |
| 98 | + TensorShape(8U), |
| 99 | + TensorShape(3U, 3U), |
| 100 | + TensorShape(2U, 5U, 5U), |
| 101 | + TensorShape(4U, 2U, 2U, 9U)}), |
| 102 | + framework::dataset::make("TensorType", { |
| 103 | + DataType::U8, |
| 104 | + DataType::S8, |
| 105 | + DataType::U32, |
| 106 | + DataType::S32, |
| 107 | + DataType::F16, |
| 108 | + DataType::F32}) |
| 109 | + ), shape, type) |
| 110 | +{ |
| 111 | + const auto t_info = TensorInfo(shape, 1, type, DataLayout::NCHW); |
| 112 | + auto t = create_tensor<Tensor>(t_info); |
| 113 | + auto t_zero = create_tensor<Tensor>(t_info); |
| 114 | + |
| 115 | + t.allocator()->allocate(); |
| 116 | + library->fill_tensor_sparse_random(Accessor(t), 0.2); |
| 117 | + |
| 118 | + t_zero.allocator()->allocate(); |
| 119 | + library->fill_static_values(Accessor(t_zero), std::vector<unsigned>(shape.total_size(), 0)); |
| 120 | + |
| 121 | + for(size_t sparse_dim = 1; sparse_dim <= shape.num_dimensions(); sparse_dim++) |
| 122 | + { |
| 123 | + auto st = t.to_coo_sparse(sparse_dim); |
| 124 | + bool is_sparse = st->info()->is_sparse(); |
| 125 | + bool is_coo = st->info()->tensor_format() == TensorFormat::COO; |
| 126 | + size_t dense_dim = shape.num_dimensions() - sparse_dim; |
| 127 | + size_t is_hybrid = dense_dim > 0; |
| 128 | + auto td = st->to_dense(); |
| 129 | + |
| 130 | + ARM_COMPUTE_EXPECT(is_sparse, framework::LogLevel::ERRORS); |
| 131 | + ARM_COMPUTE_EXPECT(is_coo, framework::LogLevel::ERRORS); |
| 132 | + ARM_COMPUTE_EXPECT(st->sparse_dim() == sparse_dim, framework::LogLevel::ERRORS); |
| 133 | + ARM_COMPUTE_EXPECT(st->dense_dim() == dense_dim, framework::LogLevel::ERRORS); |
| 134 | + ARM_COMPUTE_EXPECT(st->is_hybrid() == is_hybrid, framework::LogLevel::ERRORS); |
| 135 | + ARM_COMPUTE_EXPECT(tensors_are_equal(Accessor(t), Accessor(*td)), framework::LogLevel::ERRORS); |
| 136 | + |
| 137 | + auto st_zero = t_zero.to_coo_sparse(sparse_dim); |
| 138 | + auto td_zero = st_zero->to_dense(); |
| 139 | + ARM_COMPUTE_EXPECT(tensors_are_equal(Accessor(t_zero), Accessor(*td_zero)), framework::LogLevel::ERRORS); |
| 140 | + } |
| 141 | +} |
| 142 | +// clang-format on |
| 143 | +// *INDENT-ON* |
| 144 | + |
| 145 | +// clang-format off |
| 146 | +/** Validates TensorInfo Autopadding */ |
| 147 | +DATA_TEST_CASE(ConvertCSRTensorToDense, framework::DatasetMode::ALL, combine( |
| 148 | + framework::dataset::make("TensorShape", { |
| 149 | + TensorShape(8U), |
| 150 | + TensorShape(3U, 3U), |
| 151 | + TensorShape(2U, 5U, 5U), |
| 152 | + TensorShape(4U, 2U, 2U, 9U)}), |
| 153 | + framework::dataset::make("TensorType", { |
| 154 | + DataType::U8, |
| 155 | + DataType::S8, |
| 156 | + DataType::U32, |
| 157 | + DataType::S32, |
| 158 | + DataType::F16, |
| 159 | + DataType::F32}) |
| 160 | + ), shape, type) |
| 161 | +{ |
| 162 | + // Currently, CSRTensor only supports 2D tensors |
| 163 | + if(shape.num_dimensions() < 2) |
| 164 | + { |
| 165 | + return; |
| 166 | + } |
| 167 | + const TensorShape tensor_shape(shape[0], shape[1]); |
| 168 | + |
| 169 | + const auto t_info = TensorInfo(tensor_shape, 1, type, DataLayout::NCHW); |
| 170 | + auto t = create_tensor<Tensor>(t_info); |
| 171 | + auto t_zero = create_tensor<Tensor>(t_info); |
| 172 | + |
| 173 | + t.allocator()->allocate(); |
| 174 | + library->fill_tensor_sparse_random(Accessor(t), 0.2); |
| 175 | + |
| 176 | + t_zero.allocator()->allocate(); |
| 177 | + library->fill_static_values(Accessor(t_zero), std::vector<unsigned>(tensor_shape.total_size(), 0)); |
| 178 | + |
| 179 | + auto st = t.to_csr_sparse(); |
| 180 | + auto td = st->to_dense(); |
| 181 | + bool is_sparse = st->info()->is_sparse(); |
| 182 | + bool is_csr = st->info()->tensor_format() == TensorFormat::CSR; |
| 183 | + size_t sparse_dim = tensor_shape.num_dimensions(); |
| 184 | + size_t dense_dim = tensor_shape.num_dimensions() - sparse_dim; |
| 185 | + size_t is_hybrid = dense_dim > 0; |
| 186 | + |
| 187 | + ARM_COMPUTE_EXPECT(is_sparse, framework::LogLevel::ERRORS); |
| 188 | + ARM_COMPUTE_EXPECT(is_csr, framework::LogLevel::ERRORS); |
| 189 | + ARM_COMPUTE_EXPECT(st->sparse_dim() == sparse_dim, framework::LogLevel::ERRORS); |
| 190 | + ARM_COMPUTE_EXPECT(st->dense_dim() == dense_dim, framework::LogLevel::ERRORS); |
| 191 | + ARM_COMPUTE_EXPECT(st->is_hybrid() == is_hybrid, framework::LogLevel::ERRORS); |
| 192 | + ARM_COMPUTE_EXPECT(tensors_are_equal(Accessor(t), Accessor(*td)), framework::LogLevel::ERRORS); |
| 193 | + |
| 194 | + auto st_zero = t_zero.to_coo_sparse(sparse_dim); |
| 195 | + auto td_zero = st_zero->to_dense(); |
| 196 | + ARM_COMPUTE_EXPECT(tensors_are_equal(Accessor(t_zero), Accessor(*td_zero)), framework::LogLevel::ERRORS); |
| 197 | +} |
| 198 | +// clang-format on |
| 199 | +// *INDENT-ON* |
| 200 | + |
| 201 | +TEST_SUITE_END() // SparseTensor |
| 202 | +TEST_SUITE_END() // UNIT |
| 203 | + |
| 204 | +} // namespace validation |
| 205 | +} // namespace test |
| 206 | +} // namespace arm_compute |
0 commit comments