diff options
author | alexander <alexander.efremov@arm.com> | 2021-03-26 21:42:19 +0000 |
---|---|---|
committer | Kshitij Sisodia <kshitij.sisodia@arm.com> | 2021-03-29 16:29:55 +0100 |
commit | 3c79893217bc632c9b0efa815091bef3c779490c (patch) | |
tree | ad06b444557eb8124652b45621d736fa1b92f65d /tests/use_case/kws_asr | |
parent | 6ad6d55715928de72979b04194da1bdf04a4c51b (diff) | |
download | ml-embedded-evaluation-kit-3c79893217bc632c9b0efa815091bef3c779490c.tar.gz |
Opensource ML embedded evaluation kit21.03
Change-Id: I12e807f19f5cacad7cef82572b6dd48252fd61fd
Diffstat (limited to 'tests/use_case/kws_asr')
-rw-r--r-- | tests/use_case/kws_asr/InferenceTestDSCNN.cc | 111 | ||||
-rw-r--r-- | tests/use_case/kws_asr/InferenceTestWav2Letter.cc | 114 | ||||
-rw-r--r-- | tests/use_case/kws_asr/InitModels.cc | 52 | ||||
-rw-r--r-- | tests/use_case/kws_asr/KwsAsrTests.cc | 18 | ||||
-rw-r--r-- | tests/use_case/kws_asr/MfccTests.cc | 156 | ||||
-rw-r--r-- | tests/use_case/kws_asr/Wav2LetterPostprocessingTest.cc | 194 | ||||
-rw-r--r-- | tests/use_case/kws_asr/Wav2LetterPreprocessingTest.cc | 152 |
7 files changed, 797 insertions, 0 deletions
diff --git a/tests/use_case/kws_asr/InferenceTestDSCNN.cc b/tests/use_case/kws_asr/InferenceTestDSCNN.cc new file mode 100644 index 0000000..f0e5c02 --- /dev/null +++ b/tests/use_case/kws_asr/InferenceTestDSCNN.cc @@ -0,0 +1,111 @@ +/* + * Copyright (c) 2021 Arm Limited. All rights reserved. + * SPDX-License-Identifier: Apache-2.0 + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#include "DsCnnModel.hpp" +#include "hal.h" +#include "TestData_kws.hpp" +#include "TensorFlowLiteMicro.hpp" + +#include <catch.hpp> +#include <random> + +namespace arm { +namespace app { +namespace kws { +bool RunInference(arm::app::Model& model, const int8_t vec[]) +{ + TfLiteTensor* inputTensor = model.GetInputTensor(0); + REQUIRE(inputTensor); + + const size_t copySz = inputTensor->bytes < IFM_DATA_SIZE ? + inputTensor->bytes : + IFM_DATA_SIZE; + memcpy(inputTensor->data.data, vec, copySz); + + return model.RunInference(); +} + +bool RunInferenceRandom(arm::app::Model& model) +{ + TfLiteTensor* inputTensor = model.GetInputTensor(0); + REQUIRE(inputTensor); + + std::random_device rndDevice; + std::mt19937 mersenneGen{rndDevice()}; + std::uniform_int_distribution<short> dist {-128, 127}; + + auto gen = [&dist, &mersenneGen](){ + return dist(mersenneGen); + }; + + std::vector<int8_t> randomAudio(inputTensor->bytes); + std::generate(std::begin(randomAudio), std::end(randomAudio), gen); + + REQUIRE(RunInference(model, randomAudio.data())); + return true; +} + +template<typename T> +void TestInference(const T* input_goldenFV, const T* output_goldenFV, arm::app::Model& model) +{ + REQUIRE(RunInference(model, input_goldenFV)); + + TfLiteTensor* outputTensor = model.GetOutputTensor(0); + + REQUIRE(outputTensor); + REQUIRE(outputTensor->bytes == OFM_DATA_SIZE); + auto tensorData = tflite::GetTensorData<T>(outputTensor); + REQUIRE(tensorData); + + for (size_t i = 0; i < outputTensor->bytes; i++) { + REQUIRE((int)tensorData[i] == (int)((T)output_goldenFV[i])); + } +} + +TEST_CASE("Running random inference with Tflu and DsCnnModel Int8", "[DS_CNN]") +{ + arm::app::DsCnnModel model{}; + + REQUIRE_FALSE(model.IsInited()); + REQUIRE(model.Init()); + REQUIRE(model.IsInited()); + + REQUIRE(RunInferenceRandom(model)); +} + +TEST_CASE("Running inference with Tflu and DsCnnModel Uint8", "[DS_CNN]") +{ + for (uint32_t i = 0 ; i < NUMBER_OF_FM_FILES; ++i) { + const int8_t* input_goldenFV = get_ifm_data_array(i); + const int8_t* output_goldenFV = get_ofm_data_array(i); + + DYNAMIC_SECTION("Executing inference with re-init") + { + arm::app::DsCnnModel model{}; + + REQUIRE_FALSE(model.IsInited()); + REQUIRE(model.Init()); + REQUIRE(model.IsInited()); + + TestInference<int8_t>(input_goldenFV, output_goldenFV, model); + + } + } +} + +} //namespace +} //namespace +} //namespace diff --git a/tests/use_case/kws_asr/InferenceTestWav2Letter.cc b/tests/use_case/kws_asr/InferenceTestWav2Letter.cc new file mode 100644 index 0000000..ee63c2f --- /dev/null +++ b/tests/use_case/kws_asr/InferenceTestWav2Letter.cc @@ -0,0 +1,114 @@ +/* + * Copyright (c) 2021 Arm Limited. All rights reserved. + * SPDX-License-Identifier: Apache-2.0 + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#include "hal.h" +#include "TensorFlowLiteMicro.hpp" +#include "Wav2LetterModel.hpp" +#include "TestData_asr.hpp" + +#include <catch.hpp> +#include <random> + +namespace arm { +namespace app { +namespace asr { + +bool RunInference(arm::app::Model& model, const int8_t vec[], const size_t copySz) +{ + TfLiteTensor* inputTensor = model.GetInputTensor(0); + REQUIRE(inputTensor); + + memcpy(inputTensor->data.data, vec, copySz); + + return model.RunInference(); +} + +bool RunInferenceRandom(arm::app::Model& model) +{ + TfLiteTensor* inputTensor = model.GetInputTensor(0); + REQUIRE(inputTensor); + + std::random_device rndDevice; + std::mt19937 mersenneGen{rndDevice()}; + std::uniform_int_distribution<short> dist {-128, 127}; + + auto gen = [&dist, &mersenneGen](){ + return dist(mersenneGen); + }; + + std::vector<int8_t> randomAudio(inputTensor->bytes); + std::generate(std::begin(randomAudio), std::end(randomAudio), gen); + + REQUIRE(RunInference(model, randomAudio.data(), inputTensor->bytes)); + return true; +} + +/* Skip this test, Wav2LetterModel if not Vela optimized but only from ML-zoo will fail. */ +TEST_CASE("Running random inference with Tflu and Wav2LetterModel Int8", "[Wav2Letter][.]") +{ + arm::app::Wav2LetterModel model{}; + + REQUIRE_FALSE(model.IsInited()); + REQUIRE(model.Init()); + REQUIRE(model.IsInited()); + + REQUIRE(RunInferenceRandom(model)); +} + + +template<typename T> +void TestInference(const T* input_goldenFV, const T* output_goldenFV, arm::app::Model& model) +{ + TfLiteTensor* inputTensor = model.GetInputTensor(0); + REQUIRE(inputTensor); + + REQUIRE(RunInference(model, input_goldenFV, inputTensor->bytes)); + + TfLiteTensor* outputTensor = model.GetOutputTensor(0); + + REQUIRE(outputTensor); + REQUIRE(outputTensor->bytes == OFM_DATA_SIZE); + auto tensorData = tflite::GetTensorData<T>(outputTensor); + REQUIRE(tensorData); + + for (size_t i = 0; i < outputTensor->bytes; i++) { + REQUIRE((int)tensorData[i] == (int)((T)output_goldenFV[i])); + } +} + +TEST_CASE("Running inference with Tflu and Wav2LetterModel Int8", "[Wav2Letter][.]") +{ + for (uint32_t i = 0 ; i < NUMBER_OF_FM_FILES; ++i) { + auto input_goldenFV = get_ifm_data_array(i);; + auto output_goldenFV = get_ofm_data_array(i); + + DYNAMIC_SECTION("Executing inference with re-init") + { + arm::app::Wav2LetterModel model{}; + + REQUIRE_FALSE(model.IsInited()); + REQUIRE(model.Init()); + REQUIRE(model.IsInited()); + + TestInference<int8_t>(input_goldenFV, output_goldenFV, model); + + } + } +} + +} //namespace +} //namespace +} //namespace diff --git a/tests/use_case/kws_asr/InitModels.cc b/tests/use_case/kws_asr/InitModels.cc new file mode 100644 index 0000000..770944d --- /dev/null +++ b/tests/use_case/kws_asr/InitModels.cc @@ -0,0 +1,52 @@ +/* + * Copyright (c) 2021 Arm Limited. All rights reserved. + * SPDX-License-Identifier: Apache-2.0 + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#include "DsCnnModel.hpp" +#include "Wav2LetterModel.hpp" + +#include <catch.hpp> + +/* Skip this test, Wav2LetterModel if not Vela optimized but only from ML-zoo will fail. */ +TEST_CASE("Init two Models", "[.]") +{ + arm::app::DsCnnModel model1; + arm::app::DsCnnModel model2; + + /* Ideally we should load the wav2letter model here, but there is + * none available to run on native (ops not supported on unoptimised + * version). However, we can certainly create two instances of the + * same type of model to see if our tensor arena re-use works as + * intended. + * + * @TODO: uncomment this when this model can run on native pipeline. */ + //arm::app::Wav2LetterModel model2; /* model2. */ + + /* Load/initialise the first model. */ + REQUIRE(model1.Init()); + + /* Allocator instance should have been created. */ + REQUIRE(nullptr != model1.GetAllocator()); + + /* Load the second model using the same allocator as model 1. */ + REQUIRE(model2.Init(model1.GetAllocator())); + + /* Make sure they point to the same allocator object. */ + REQUIRE(model1.GetAllocator() == model2.GetAllocator()); + + /* Both models should report being initialised. */ + REQUIRE(true == model1.IsInited()); + REQUIRE(true == model2.IsInited()); +}
\ No newline at end of file diff --git a/tests/use_case/kws_asr/KwsAsrTests.cc b/tests/use_case/kws_asr/KwsAsrTests.cc new file mode 100644 index 0000000..09f82da --- /dev/null +++ b/tests/use_case/kws_asr/KwsAsrTests.cc @@ -0,0 +1,18 @@ +/* + * Copyright (c) 2021 Arm Limited. All rights reserved. + * SPDX-License-Identifier: Apache-2.0 + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#define CATCH_CONFIG_MAIN +#include <catch.hpp> diff --git a/tests/use_case/kws_asr/MfccTests.cc b/tests/use_case/kws_asr/MfccTests.cc new file mode 100644 index 0000000..9509519 --- /dev/null +++ b/tests/use_case/kws_asr/MfccTests.cc @@ -0,0 +1,156 @@ +/* + * Copyright (c) 2021 Arm Limited. All rights reserved. + * SPDX-License-Identifier: Apache-2.0 + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#include "DsCnnMfcc.hpp" + +#include <algorithm> +#include <catch.hpp> +#include <limits> + +/* First 640 samples from yes.wav. */ +const std::vector<int16_t> testWav = std::vector<int16_t>{ + 139, 143, 164, 163, 157, 156, 151, 148, 172, 171, + 165, 169, 149, 142, 145, 147, 166, 146, 112, 132, + 132, 136, 165, 176, 176, 152, 138, 158, 179, 185, + 183, 148, 121, 130, 167, 204, 163, 132, 165, 184, + 193, 205, 210, 204, 195, 178, 168, 197, 207, 201, + 197, 177, 185, 196, 191, 198, 196, 183, 193, 181, + 157, 170, 167, 159, 164, 152, 146, 167, 180, 171, + 194, 232, 204, 173, 171, 172, 184, 169, 175, 199, + 200, 195, 185, 214, 214, 193, 196, 191, 204, 191, + 172, 187, 183, 192, 203, 172, 182, 228, 232, 205, + 177, 174, 191, 210, 210, 211, 197, 177, 198, 217, + 233, 236, 203, 191, 169, 145, 149, 161, 198, 206, + 176, 137, 142, 181, 200, 215, 201, 188, 166, 162, + 184, 155, 135, 132, 126, 142, 169, 184, 172, 156, + 132, 119, 150, 147, 154, 160, 125, 130, 137, 154, + 161, 168, 195, 182, 160, 134, 138, 146, 130, 120, + 101, 122, 137, 118, 117, 131, 145, 140, 146, 148, + 148, 168, 159, 134, 114, 114, 130, 147, 147, 134, + 125, 98, 107, 127, 99, 79, 84, 107, 117, 114, + 93, 92, 127, 112, 109, 110, 96, 118, 97, 87, + 110, 95, 128, 153, 147, 165, 146, 106, 101, 137, + 139, 96, 73, 90, 91, 51, 69, 102, 100, 103, + 96, 101, 123, 107, 82, 89, 118, 127, 99, 100, + 111, 97, 111, 123, 106, 121, 133, 103, 100, 88, + 85, 111, 114, 125, 102, 91, 97, 84, 139, 157, + 109, 66, 72, 129, 111, 90, 127, 126, 101, 109, + 142, 138, 129, 159, 140, 80, 74, 78, 76, 98, + 68, 42, 106, 143, 112, 102, 115, 114, 82, 75, + 92, 80, 110, 114, 66, 86, 119, 101, 101, 103, + 118, 145, 85, 40, 62, 88, 95, 87, 73, 64, + 86, 71, 71, 105, 80, 73, 96, 92, 85, 90, + 81, 86, 105, 100, 89, 78, 102, 114, 95, 98, + 69, 70, 108, 112, 111, 90, 104, 137, 143, 160, + 145, 121, 98, 86, 91, 87, 115, 123, 109, 99, + 85, 120, 131, 116, 125, 144, 153, 111, 98, 110, + 93, 89, 101, 137, 155, 142, 108, 94, 136, 145, + 129, 129, 122, 109, 90, 76, 81, 110, 119, 96, + 95, 102, 105, 111, 90, 89, 111, 115, 86, 51, + 107, 140, 105, 105, 110, 142, 125, 76, 75, 69, + 65, 52, 61, 69, 55, 42, 47, 58, 37, 35, + 24, 20, 44, 22, 16, 26, 6, 3, 4, 23, + 60, 51, 30, 12, 24, 31, -9, -16, -13, 13, + 19, 9, 37, 55, 70, 36, 23, 57, 45, 33, + 50, 59, 18, 11, 62, 74, 52, 8, -3, 26, + 51, 48, -5, -9, 12, -7, -12, -5, 28, 41, + -2, -30, -13, 31, 33, -12, -22, -8, -15, -17, + 2, -6, -25, -27, -24, -8, 4, -9, -52, -47, + -9, -32, -45, -5, 41, 15, -32, -14, 2, -1, + -10, -30, -32, -25, -21, -17, -14, 8, -4, -13, + 34, 18, -36, -38, -18, -19, -28, -17, -14, -16, + -2, -20, -27, 12, 11, -17, -33, -12, -22, -64, + -42, -26, -23, -22, -37, -51, -53, -30, -18, -48, + -69, -38, -54, -96, -72, -49, -50, -57, -41, -22, + -43, -64, -54, -23, -49, -69, -41, -44, -42, -49, + -40, -26, -54, -50, -38, -49, -70, -94, -89, -69, + -56, -65, -71, -47, -39, -49, -79, -91, -56, -46, + -62, -86, -64, -32, -47, -50, -71, -77, -65, -68, + -52, -51, -61, -67, -61, -81, -93, -52, -59, -62, + -51, -75, -76, -50, -32, -54, -68, -70, -43, 1, + -42, -92, -80, -41, -38, -79, -69, -49, -82, -122, + -93, -21, -24, -61, -70, -73, -62, -74, -69, -43, + -25, -15, -43, -23, -26, -69, -44, -12, 1, -51, + -78, -13, 3, -53, -105, -72, -24, -62, -66, -31, + -40, -65, -86, -64, -44, -55, -63, -61, -37, -41, +}; + +/* Golden audio ops mfcc output for the above wav. */ +const std::vector<float> testWavMfcc { + -22.67135, -0.61615, 2.07233, 0.58137, 1.01655, 0.85816, 0.46039, 0.03393, 1.16511, 0.0072, +}; + +arm::app::audio::DsCnnMFCC GetMFCCInstance() { + const int sampFreq = arm::app::audio::DsCnnMFCC::ms_defaultSamplingFreq; + const int frameLenMs = 40; + const int frameLenSamples = sampFreq * frameLenMs * 0.001; + const int numMfccFeats = 10; + + return arm::app::audio::DsCnnMFCC(numMfccFeats, frameLenSamples); +} + +template <class T> +void TestQuntisedMFCC() { + const float quantScale = 1.1088106632232666; + const int quantOffset = 95; + std::vector<T> mfccOutput = GetMFCCInstance().MfccComputeQuant<T>(testWav, quantScale, quantOffset); + + const long min_val = std::numeric_limits<T>::min(); + const long max_val = std::numeric_limits<T>::max(); + + for (size_t i = 0; i < testWavMfcc.size(); ++i){ + long TestWavMfcc = (std::lround((testWavMfcc[i] / quantScale) + quantOffset)); + T quantizedTestWavMfcc = static_cast<T>(std::max(min_val, std::min(TestWavMfcc, max_val))); + + REQUIRE(quantizedTestWavMfcc == Approx(mfccOutput[i]).margin(0)); + } +} +template void TestQuntisedMFCC<int8_t>(); +template void TestQuntisedMFCC<uint8_t>(); +template void TestQuntisedMFCC<int16_t>(); + +TEST_CASE("MFCC calculation test") +{ + hal_platform platform; + data_acq_module dataAcq; + data_psn_module dataPsn; + platform_timer timer; + + /* Initialise the HAL and platform. */ + hal_init(&platform, &dataAcq, &dataPsn, &timer); + hal_platform_init(&platform); + + SECTION("FP32") + { + auto mfccOutput = GetMFCCInstance().MfccCompute(testWav); + REQUIRE_THAT( mfccOutput, Catch::Approx( testWavMfcc ).margin(0.0001) ); + } + + SECTION("int8_t") + { + TestQuntisedMFCC<int8_t>(); + } + + SECTION("uint8_t") + { + TestQuntisedMFCC<uint8_t>(); + } + + SECTION("MFCC quant calculation test - int16_t") + { + TestQuntisedMFCC<int16_t>(); + } +}
\ No newline at end of file diff --git a/tests/use_case/kws_asr/Wav2LetterPostprocessingTest.cc b/tests/use_case/kws_asr/Wav2LetterPostprocessingTest.cc new file mode 100644 index 0000000..6fd7df3 --- /dev/null +++ b/tests/use_case/kws_asr/Wav2LetterPostprocessingTest.cc @@ -0,0 +1,194 @@ +/* + * Copyright (c) 2021 Arm Limited. All rights reserved. + * SPDX-License-Identifier: Apache-2.0 + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#include "Wav2LetterPostprocess.hpp" +#include "Wav2LetterModel.hpp" + +#include <algorithm> +#include <catch.hpp> +#include <limits> + +template <typename T> +static TfLiteTensor GetTestTensor(std::vector <int>& shape, + T initVal, + std::vector<T>& vectorBuf) +{ + REQUIRE(0 != shape.size()); + + shape.insert(shape.begin(), shape.size()); + uint32_t sizeInBytes = sizeof(T); + for (size_t i = 1; i < shape.size(); ++i) { + sizeInBytes *= shape[i]; + } + + /* Allocate mem. */ + vectorBuf = std::vector<T>(sizeInBytes, initVal); + TfLiteIntArray* dims = tflite::testing::IntArrayFromInts(shape.data()); + return tflite::testing::CreateQuantizedTensor( + vectorBuf.data(), dims, + 1, 0, "test-tensor"); +} + +TEST_CASE("Checking return value") +{ + SECTION("Mismatched post processing parameters and tensor size") + { + const uint32_t ctxLen = 5; + const uint32_t innerLen = 3; + arm::app::audio::asr::Postprocess post{ctxLen, innerLen, 0}; + + std::vector <int> tensorShape = {1, 1, 1, 13}; + std::vector <int8_t> tensorVec; + TfLiteTensor tensor = GetTestTensor<int8_t>( + tensorShape, 100, tensorVec); + REQUIRE(false == post.Invoke(&tensor, arm::app::Wav2LetterModel::ms_outputRowsIdx, false)); + } + + SECTION("Post processing succeeds") + { + const uint32_t ctxLen = 5; + const uint32_t innerLen = 3; + arm::app::audio::asr::Postprocess post{ctxLen, innerLen, 0}; + + std::vector <int> tensorShape = {1, 1, 13, 1}; + std::vector <int8_t> tensorVec; + TfLiteTensor tensor = GetTestTensor<int8_t>( + tensorShape, 100, tensorVec); + + /* Copy elements to compare later. */ + std::vector <int8_t> originalVec = tensorVec; + + /* This step should not erase anything. */ + REQUIRE(true == post.Invoke(&tensor, arm::app::Wav2LetterModel::ms_outputRowsIdx, false)); + } +} + +TEST_CASE("Postprocessing - erasing required elements") +{ + constexpr uint32_t ctxLen = 5; + constexpr uint32_t innerLen = 3; + constexpr uint32_t nRows = 2*ctxLen + innerLen; + constexpr uint32_t nCols = 10; + constexpr uint32_t blankTokenIdx = nCols - 1; + std::vector <int> tensorShape = {1, 1, nRows, nCols}; + + SECTION("First and last iteration") + { + arm::app::audio::asr::Postprocess post{ctxLen, innerLen, blankTokenIdx}; + std::vector <int8_t> tensorVec; + TfLiteTensor tensor = GetTestTensor<int8_t>( + tensorShape, 100, tensorVec); + + /* Copy elements to compare later. */ + std::vector <int8_t> originalVec = tensorVec; + + /* This step should not erase anything. */ + REQUIRE(true == post.Invoke(&tensor, arm::app::Wav2LetterModel::ms_outputRowsIdx, true)); + REQUIRE(originalVec == tensorVec); + } + + SECTION("Right context erase") + { + arm::app::audio::asr::Postprocess post{ctxLen, innerLen, blankTokenIdx}; + + std::vector <int8_t> tensorVec; + TfLiteTensor tensor = GetTestTensor<int8_t>( + tensorShape, 100, tensorVec); + + /* Copy elements to compare later. */ + std::vector <int8_t> originalVec = tensorVec; + + /* This step should erase the right context only. */ + REQUIRE(true == post.Invoke(&tensor, arm::app::Wav2LetterModel::ms_outputRowsIdx, false)); + REQUIRE(originalVec != tensorVec); + + /* The last ctxLen * 10 elements should be gone. */ + for (size_t i = 0; i < ctxLen; ++i) { + for (size_t j = 0; j < nCols; ++j) { + /* Check right context elements are zeroed. */ + if (j == blankTokenIdx) { + CHECK(tensorVec[(ctxLen + innerLen) * nCols + i*nCols + j] == 1); + } else { + CHECK(tensorVec[(ctxLen + innerLen) * nCols + i*nCols + j] == 0); + } + + /* Check left context is preserved. */ + CHECK(tensorVec[i*nCols + j] == originalVec[i*nCols + j]); + } + } + + /* Check inner elements are preserved. */ + for (size_t i = ctxLen * nCols; i < (ctxLen + innerLen) * nCols; ++i) { + CHECK(tensorVec[i] == originalVec[i]); + } + } + + SECTION("Left and right context erase") + { + arm::app::audio::asr::Postprocess post{ctxLen, innerLen, blankTokenIdx}; + + std::vector <int8_t> tensorVec; + TfLiteTensor tensor = GetTestTensor<int8_t>(tensorShape, 100, tensorVec); + + /* Copy elements to compare later. */ + std::vector <int8_t> originalVec = tensorVec; + + /* This step should erase right context. */ + REQUIRE(true == post.Invoke(&tensor, arm::app::Wav2LetterModel::ms_outputRowsIdx, false)); + + /* Calling it the second time should erase the left context. */ + REQUIRE(true == post.Invoke(&tensor, arm::app::Wav2LetterModel::ms_outputRowsIdx, false)); + + REQUIRE(originalVec != tensorVec); + + /* The first and last ctxLen * 10 elements should be gone. */ + for (size_t i = 0; i < ctxLen; ++i) { + for (size_t j = 0; j < nCols; ++j) { + /* Check left and right context elements are zeroed. */ + if (j == blankTokenIdx) { + CHECK(tensorVec[(ctxLen + innerLen) * nCols + i * nCols + j] == 1); + CHECK(tensorVec[i * nCols + j] == 1); + } else { + CHECK(tensorVec[(ctxLen + innerLen) * nCols + i * nCols + j] == 0); + CHECK(tensorVec[i * nCols + j] == 0); + } + } + } + + /* Check inner elements are preserved. */ + for (size_t i = ctxLen * nCols; i < (ctxLen + innerLen) * nCols; ++i) { + /* Check left context is preserved. */ + CHECK(tensorVec[i] == originalVec[i]); + } + } + + SECTION("Try left context erase") + { + /* Should not be able to erase the left context if it is the first iteration. */ + arm::app::audio::asr::Postprocess post{ctxLen, innerLen, blankTokenIdx}; + + std::vector <int8_t> tensorVec; + TfLiteTensor tensor = GetTestTensor<int8_t>( + tensorShape, 100, tensorVec); + + /* Copy elements to compare later. */ + std::vector <int8_t> originalVec = tensorVec; + + /* Calling it the second time should erase the left context. */ + REQUIRE(true == post.Invoke(&tensor, arm::app::Wav2LetterModel::ms_outputRowsIdx, true)); + REQUIRE(originalVec == tensorVec); + } +}
\ No newline at end of file diff --git a/tests/use_case/kws_asr/Wav2LetterPreprocessingTest.cc b/tests/use_case/kws_asr/Wav2LetterPreprocessingTest.cc new file mode 100644 index 0000000..e71366a --- /dev/null +++ b/tests/use_case/kws_asr/Wav2LetterPreprocessingTest.cc @@ -0,0 +1,152 @@ +/* + * Copyright (c) 2021 Arm Limited. All rights reserved. + * SPDX-License-Identifier: Apache-2.0 + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +#include "Wav2LetterPreprocess.hpp" + +#include <algorithm> +#include <catch.hpp> +#include <limits> + +constexpr uint32_t numMfccFeatures = 13; +constexpr uint32_t numMfccVectors = 10; + +/* Test vector output: generated using test-asr-preprocessing.py. */ +int8_t expectedResult[numMfccVectors][numMfccFeatures*3] = { + /* Feature vec 0. */ + -32, 4, -9, -8, -10, -10, -11, -11, -11, -11, -12, -11, -11, /* MFCCs. */ + -11, -11, -11, -11, -11, -11, -11, -11, -11, -11, -11, -11, -11, /* Delta 1. */ + -10, -10, -10, -10, -10, -10, -10, -10, -10, -10, -10, -10, -10, /* Delta 2. */ + + /* Feature vec 1. */ + -31, 4, -9, -8, -10, -10, -11, -11, -11, -11, -12, -11, -11, + -11, -11, -11, -11, -11, -11, -11, -11, -11, -11, -11, -11, -11, + -10, -10, -10, -10, -10, -10, -10, -10, -10, -10, -10, -10, -10, + + /* Feature vec 2. */ + -31, 4, -9, -9, -10, -10, -11, -11, -11, -11, -12, -12, -12, + -11, -11, -11, -11, -11, -11, -11, -11, -11, -11, -11, -11, -11, + -10, -10, -10, -10, -10, -10, -10, -10, -10, -10, -10, -10, -10, + + /* Feature vec 3. */ + -31, 4, -9, -9, -10, -10, -11, -11, -11, -11, -11, -12, -12, + -11, -11, -11, -11, -11, -11, -11, -11, -11, -11, -11, -11, -11, + -10, -10, -10, -10, -10, -10, -10, -10, -10, -10, -10, -10, -10, + + /* Feature vec 4 : this should have valid delta 1 and delta 2. */ + -31, 4, -9, -9, -10, -10, -11, -11, -11, -11, -11, -12, -12, + -38, -29, -9, 1, -2, -7, -8, -8, -12, -16, -14, -5, 5, + -68, -50, -13, 5, 0, -9, -9, -8, -13, -20, -19, -3, 15, + + /* Feature vec 5 : this should have valid delta 1 and delta 2. */ + -31, 4, -9, -8, -10, -10, -11, -11, -11, -11, -11, -12, -12, + -62, -45, -11, 5, 0, -8, -9, -8, -12, -19, -17, -3, 13, + -27, -22, -13, -9, -11, -12, -12, -11, -11, -13, -13, -10, -6, + + /* Feature vec 6. */ + -31, 4, -9, -8, -10, -10, -11, -11, -11, -11, -12, -11, -11, + -11, -11, -11, -11, -11, -11, -11, -11, -11, -11, -11, -11, -11, + -10, -10, -10, -10, -10, -10, -10, -10, -10, -10, -10, -10, -10, + + /* Feature vec 7. */ + -32, 4, -9, -8, -10, -10, -11, -11, -11, -12, -12, -11, -11, + -11, -11, -11, -11, -11, -11, -11, -11, -11, -11, -11, -11, -11, + -10, -10, -10, -10, -10, -10, -10, -10, -10, -10, -10, -10, -10, + + /* Feature vec 8. */ + -32, 4, -9, -8, -10, -10, -11, -11, -11, -12, -12, -11, -11, + -11, -11, -11, -11, -11, -11, -11, -11, -11, -11, -11, -11, -11, + -10, -10, -10, -10, -10, -10, -10, -10, -10, -10, -10, -10, -10, + + /* Feature vec 9. */ + -31, 4, -9, -8, -10, -10, -11, -11, -11, -11, -12, -11, -11, + -11, -11, -11, -11, -11, -11, -11, -11, -11, -11, -11, -11, -11, + -10, -10, -10, -10, -10, -10, -10, -10, -10, -10, -10, -10, -10 +}; + +void PopulateTestWavVector(std::vector<int16_t>& vec) +{ + constexpr int int16max = std::numeric_limits<int16_t>::max(); + int val = 0; + for (size_t i = 0; i < vec.size(); ++i, ++val) { + + /* We want a differential filter response from both - order 1 + * and 2 => Don't have a linear signal here - we use a signal + * using squares for example. Alternate sign flips might work + * just as well and will be computationally less work! */ + int valsq = val * val; + if (valsq > int16max) { + val = 0; + valsq = 0; + } + vec[i] = valsq; + } +} + +TEST_CASE("Preprocessing calculation INT8") +{ + /* Initialise the HAL and platform. */ + hal_platform platform; + data_acq_module data_acq; + data_psn_module data_psn; + platform_timer timer; + hal_init(&platform, &data_acq, &data_psn, &timer); + hal_platform_init(&platform); + + /* Constants. */ + const uint32_t windowLen = 512; + const uint32_t windowStride = 160; + const int dimArray[] = {3, 1, numMfccFeatures * 3, numMfccVectors}; + const float quantScale = 0.1410219967365265; + const int quantOffset = -11; + + /* Test wav memory. */ + std::vector <int16_t> testWav((windowStride * numMfccVectors) + + (windowLen - windowStride)); + + /* Populate with dummy input. */ + PopulateTestWavVector(testWav); + + /* Allocate mem for tensor. */ + std::vector<int8_t> tensorVec(dimArray[1]*dimArray[2]*dimArray[3]); + + /* Initialise dimensions and the test tensor. */ + TfLiteIntArray* dims= tflite::testing::IntArrayFromInts(dimArray); + TfLiteTensor tensor = tflite::testing::CreateQuantizedTensor( + tensorVec.data(), dims, quantScale, quantOffset, "preprocessedInput"); + + /* Initialise pre-processing module. */ + arm::app::audio::asr::Preprocess prep{ + numMfccFeatures, windowLen, windowStride, numMfccVectors}; + + /* Invoke pre-processing. */ + REQUIRE(prep.Invoke(testWav.data(), testWav.size(), &tensor)); + + /* Wrap the tensor with a std::vector for ease. */ + int8_t * tensorData = tflite::GetTensorData<int8_t>(&tensor); + std::vector <int8_t> vecResults = + std::vector<int8_t>(tensorData, tensorData + tensor.bytes); + + /* Check sizes. */ + REQUIRE(vecResults.size() == sizeof(expectedResult)); + + /* Check that the elements have been calculated correctly. */ + for (uint32_t j = 0; j < numMfccVectors; ++j) { + for (uint32_t i = 0; i < numMfccFeatures * 3; ++i) { + size_t tensorIdx = (j * numMfccFeatures * 3) + i; + CHECK(vecResults[tensorIdx] == expectedResult[j][i]); + } + } +} |