1 /*
2 * Copyright (c) 2018-2021 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/runtime/NEON/functions/NELSTMLayer.h"
25 #include "tests/NEON/Accessor.h"
26 #include "tests/PaddingCalculator.h"
27 #include "tests/datasets/LSTMLayerDataset.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/fixtures/LSTMLayerFixture.h"
33
34 namespace arm_compute
35 {
36 namespace test
37 {
38 namespace validation
39 {
40 namespace
41 {
42 RelativeTolerance<float> tolerance_f32(0.00001f);
43 RelativeTolerance<half> tolerance_f16(half(0.1));
44 } // namespace
45
46 TEST_SUITE(NEON)
TEST_SUITE(LSTMLayer)47 TEST_SUITE(LSTMLayer)
48
49 // *INDENT-OFF*
50 // clang-format off
51 DATA_TEST_CASE(Validate, framework::DatasetMode::ALL, zip(zip(zip(zip(zip(zip(zip(zip(zip(
52 framework::dataset::make("InputInfo", { TensorInfo(TensorShape(8U, 2U), 1, DataType::U8), // Wrong data type
53 TensorInfo(TensorShape(27U, 13U, 2U), 1, DataType::F32), // Wrong input size
54 TensorInfo(TensorShape(8U, 2U), 1, DataType::F32), // Wrong input weights size
55 TensorInfo(TensorShape(8U, 2U), 1, DataType::F32), // Wrong recurrent weights size
56 TensorInfo(TensorShape(8U, 2U), 1, DataType::F32), // Wrong cell bias size
57 TensorInfo(TensorShape(8U, 2U), 1, DataType::F32), // Wrong cell state size
58 TensorInfo(TensorShape(8U, 2U), 1, DataType::F32), // Wrong output size
59 TensorInfo(TensorShape(8U, 2U), 1, DataType::F32), // Wrong scratch size
60 }),
61 framework::dataset::make("InputWeightsInfo", { TensorInfo(TensorShape(8U, 16U), 1, DataType::F32),
62 TensorInfo(TensorShape(8U, 16U), 1, DataType::F32),
63 TensorInfo(TensorShape(27U, 11U, 2U), 1, DataType::F32),
64 TensorInfo(TensorShape(8U, 16U), 1, DataType::F32),
65 TensorInfo(TensorShape(8U, 16U), 1, DataType::F32),
66 TensorInfo(TensorShape(8U, 16U), 1, DataType::F32),
67 TensorInfo(TensorShape(8U, 16U), 1, DataType::F32),
68 TensorInfo(TensorShape(8U, 16U), 1, DataType::F32),
69 })),
70 framework::dataset::make("RecurrentWeightsInfo", { TensorInfo(TensorShape(16U, 16U), 1, DataType::F32),
71 TensorInfo(TensorShape(16U, 16U), 1, DataType::F32),
72 TensorInfo(TensorShape(16U, 16U), 1, DataType::F32),
73 TensorInfo(TensorShape(25U, 11U, 2U), 1, DataType::F32),
74 TensorInfo(TensorShape(16U, 16U), 1, DataType::F32),
75 TensorInfo(TensorShape(16U, 16U), 1, DataType::F32),
76 TensorInfo(TensorShape(16U, 16U), 1, DataType::F32),
77 TensorInfo(TensorShape(16U, 16U), 1, DataType::F32),
78 })),
79 framework::dataset::make("CellBiasInfo", { TensorInfo(TensorShape(16U), 1, DataType::F32),
80 TensorInfo(TensorShape(16U), 1, DataType::F32),
81 TensorInfo(TensorShape(16U), 1, DataType::F32),
82 TensorInfo(TensorShape(16U), 1, DataType::F32),
83 TensorInfo(TensorShape(30U), 1, DataType::F32),
84 TensorInfo(TensorShape(16U), 1, DataType::F32),
85 TensorInfo(TensorShape(16U), 1, DataType::F32),
86 TensorInfo(TensorShape(16U), 1, DataType::F32),
87 })),
88 framework::dataset::make("ProjectionBiasInfo", { TensorInfo(TensorShape(16U), 1, DataType::F32),
89 TensorInfo(TensorShape(16U), 1, DataType::F32),
90 TensorInfo(TensorShape(16U), 1, DataType::F32),
91 TensorInfo(TensorShape(16U), 1, DataType::F32),
92 TensorInfo(TensorShape(16U), 1, DataType::F32),
93 TensorInfo(TensorShape(16U), 1, DataType::F32),
94 TensorInfo(TensorShape(16U), 1, DataType::F32),
95 TensorInfo(TensorShape(16U), 1, DataType::F32),
96 })),
97 framework::dataset::make("CellStateInfo", { TensorInfo(TensorShape(16U, 2U), 1, DataType::F32),
98 TensorInfo(TensorShape(16U, 2U), 1, DataType::F32),
99 TensorInfo(TensorShape(16U, 2U), 1, DataType::F32),
100 TensorInfo(TensorShape(16U, 2U), 1, DataType::F32),
101 TensorInfo(TensorShape(16U, 2U), 1, DataType::F32),
102 TensorInfo(TensorShape(11U), 1, DataType::F32),
103 TensorInfo(TensorShape(16U, 2U), 1, DataType::F32),
104 TensorInfo(TensorShape(16U, 2U), 1, DataType::F32),
105 })),
106 framework::dataset::make("OutputInfo", { TensorInfo(TensorShape(16U, 2U), 1, DataType::F32),
107 TensorInfo(TensorShape(16U, 2U), 1, DataType::F32),
108 TensorInfo(TensorShape(16U, 2U), 1, DataType::F32),
109 TensorInfo(TensorShape(16U, 2U), 1, DataType::F32),
110 TensorInfo(TensorShape(16U, 2U), 1, DataType::F32),
111 TensorInfo(TensorShape(16U, 2U), 1, DataType::F32),
112 TensorInfo(TensorShape(11U, 2U), 1, DataType::F32),
113 TensorInfo(TensorShape(16U, 2U), 1, DataType::F32),
114 })),
115 framework::dataset::make("ScratchInfo", { TensorInfo(TensorShape(64U, 2U), 1, DataType::F32),
116 TensorInfo(TensorShape(64U, 2U), 1, DataType::F32),
117 TensorInfo(TensorShape(64U, 2U), 1, DataType::F32),
118 TensorInfo(TensorShape(64U, 2U), 1, DataType::F32),
119 TensorInfo(TensorShape(64U, 2U), 1, DataType::F32),
120 TensorInfo(TensorShape(64U, 2U), 1, DataType::F32),
121 TensorInfo(TensorShape(64U, 2U), 1, DataType::F32),
122 TensorInfo(TensorShape(12U, 2U), 1, DataType::F32),
123 })),
124 framework::dataset::make("ActivationInfo", { ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU),
125 ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU),
126 ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU),
127 ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU),
128 ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU),
129 ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU),
130 ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU),
131 ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU),
132 })),
133 framework::dataset::make("Expected", { false, false, false, false, false, false, false, false })),
134 input_info, input_weights_info, recurrent_weights_info, cell_bias_info, projection_bias_info, cell_state_info, output_info, scratch_info, info, expected)
135 {
136 LSTMParams<ITensorInfo> lstm_params_info;
137 auto cell_bias_clone = cell_bias_info.clone();
138 lstm_params_info.set_peephole_params(cell_bias_clone.get(), cell_bias_clone.get())
139 .set_projection_params(&recurrent_weights_info, &projection_bias_info)
140 .set_cifg_params(&input_weights_info, &recurrent_weights_info, cell_bias_clone.get(), cell_bias_clone.get());
141
142 ARM_COMPUTE_EXPECT(bool(NELSTMLayer::validate(&input_info.clone()->set_is_resizable(false), &input_weights_info.clone()->set_is_resizable(false), &input_weights_info.clone()->set_is_resizable(false),
143 &input_weights_info.clone()->set_is_resizable(false), &recurrent_weights_info.clone()->set_is_resizable(false), &recurrent_weights_info.clone()->set_is_resizable(false),
144 &recurrent_weights_info.clone()->set_is_resizable(false), &cell_bias_info.clone()->set_is_resizable(false), &cell_bias_info.clone()->set_is_resizable(false),
145 &cell_bias_info.clone()->set_is_resizable(false),
146 &output_info.clone()->set_is_resizable(false), &cell_state_info.clone()->set_is_resizable(false),
147 &scratch_info.clone()->set_is_resizable(false), &output_info.clone()->set_is_resizable(false), &cell_state_info.clone()->set_is_resizable(false), &output_info.clone()->set_is_resizable(false),
148 lstm_params_info, info, 0.05, 0.9)) == expected, framework::LogLevel::ERRORS);
149 }
150 // clang-format on
151 // *INDENT-ON*
152
153 template <typename T>
154 using NELSTMLayerFixture = LSTMLayerValidationFixture<Tensor, Accessor, NELSTMLayer, LSTMParams<ITensor>, T>;
155
156 TEST_SUITE(FP32)
157 FIXTURE_DATA_TEST_CASE(RunSmall, NELSTMLayerFixture<float>, framework::DatasetMode::ALL, combine(combine(combine(combine(datasets::SmallLSTMLayerDataset(), framework::dataset::make("DataType",
158 DataType::F32)),
159 framework::dataset::make("ProjectionOpt", { true, false })),
160 framework::dataset::make("PeepholeOpt", { true, false })),
161 framework::dataset::make("UseLayerNorm", { true, false })))
162 {
163 // Validate output
164 validate(Accessor(_target), _reference, tolerance_f32);
165 validate(Accessor(_target_scratch), _reference_scratch, tolerance_f32);
166 }
167 TEST_SUITE_END() // FP32
168
169 #ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC
TEST_SUITE(FP16)170 TEST_SUITE(FP16)
171 FIXTURE_DATA_TEST_CASE(RunSmall, NELSTMLayerFixture<half>, framework::DatasetMode::ALL, combine(combine(combine(combine(datasets::SmallLSTMLayerDataset(), framework::dataset::make("DataType",
172 DataType::F16)),
173 framework::dataset::make("ProjectionOpt", { true, false })),
174 framework::dataset::make("PeepholeOpt", { true, false })),
175 framework::dataset::make("UseLayerNorm", { true, false })))
176 {
177 // Validate output
178 validate(Accessor(_target), _reference, tolerance_f16);
179 validate(Accessor(_target_scratch), _reference_scratch, tolerance_f16);
180 }
181 TEST_SUITE_END() // FP16
182 #endif /* __ARM_FEATURE_FP16_VECTOR_ARITHMETIC */
183 TEST_SUITE_END() // LSTMLayer
184 TEST_SUITE_END() // Neon
185 } // namespace validation
186 } // namespace test
187 } // namespace arm_compute
188