From 0d110594b8a50ce3311be5187f01de2e3b8fe995 Mon Sep 17 00:00:00 2001 From: Richard Burton Date: Thu, 12 Aug 2021 17:26:30 +0100 Subject: MLECO-1904: Update to use latest TFLu * Now uses seperate TFLu github repo * Fixes to align with API changes * Update ASR model ops and re-enable ASR inference tests * Set default release level to release_with_logs Signed-off-by: Richard Burton Change-Id: I57612088985dece1413c5c00a6e442381e07dd91 --- tests/common/ClassifierTests.cc | 2 +- tests/use_case/asr/AsrClassifierTests.cc | 4 ++-- tests/use_case/asr/InferenceTestWav2Letter.cc | 5 ++--- tests/use_case/asr/Wav2LetterPreprocessingTest.cc | 2 +- tests/use_case/img_class/InferenceTestMobilenetV2.cc | 2 +- tests/use_case/kws_asr/InferenceTestWav2Letter.cc | 5 ++--- tests/use_case/kws_asr/Wav2LetterPreprocessingTest.cc | 2 +- 7 files changed, 10 insertions(+), 12 deletions(-) (limited to 'tests') diff --git a/tests/common/ClassifierTests.cc b/tests/common/ClassifierTests.cc index a04e4c2..d950304 100644 --- a/tests/common/ClassifierTests.cc +++ b/tests/common/ClassifierTests.cc @@ -21,7 +21,7 @@ template void test_classifier_result(std::vector>& selectedResults, T defaultTensorValue) { - const int dimArray[] = {1, 1001}; + int dimArray[] = {1, 1001}; std::vector labels(1001); std::vector outputVec(1001, defaultTensorValue); TfLiteIntArray* dims= tflite::testing::IntArrayFromInts(dimArray); diff --git a/tests/use_case/asr/AsrClassifierTests.cc b/tests/use_case/asr/AsrClassifierTests.cc index 12523aa..e2bfb18 100644 --- a/tests/use_case/asr/AsrClassifierTests.cc +++ b/tests/use_case/asr/AsrClassifierTests.cc @@ -30,7 +30,7 @@ TEST_CASE("Test invalid classifier") TEST_CASE("Test valid classifier UINT8") { - const int dimArray[] = {4, 1, 1, 246, 29}; + int dimArray[] = {4, 1, 1, 246, 29}; std::vector labels(29); std::vector outputVec(7134); TfLiteIntArray* dims= tflite::testing::IntArrayFromInts(dimArray); @@ -46,7 +46,7 @@ TEST_CASE("Test valid classifier UINT8") { TEST_CASE("Get classification results") { - const int dimArray[] = {4, 1, 1, 10, 15}; + int dimArray[] = {4, 1, 1, 10, 15}; std::vector labels(15); std::vector outputVec(150, static_cast(1)); TfLiteIntArray* dims= tflite::testing::IntArrayFromInts(dimArray); diff --git a/tests/use_case/asr/InferenceTestWav2Letter.cc b/tests/use_case/asr/InferenceTestWav2Letter.cc index 0943db8..d5e6c35 100644 --- a/tests/use_case/asr/InferenceTestWav2Letter.cc +++ b/tests/use_case/asr/InferenceTestWav2Letter.cc @@ -54,8 +54,7 @@ bool RunInferenceRandom(arm::app::Model& model) return true; } -/* Skip this test, Wav2LetterModel if not Vela optimized but only from ML-zoo will fail. */ -TEST_CASE("Running random inference with TensorFlow Lite Micro and Wav2LetterModel Int8", "[Wav2Letter][.]") +TEST_CASE("Running random inference with TensorFlow Lite Micro and Wav2LetterModel Int8", "[Wav2Letter]") { arm::app::Wav2LetterModel model{}; @@ -86,7 +85,7 @@ void TestInference(const T* input_goldenFV, const T* output_goldenFV, arm::app:: } } -TEST_CASE("Running inference with Tflu and Wav2LetterModel Int8", "[Wav2Letter][.]") +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);; diff --git a/tests/use_case/asr/Wav2LetterPreprocessingTest.cc b/tests/use_case/asr/Wav2LetterPreprocessingTest.cc index 1391011..8af9014 100644 --- a/tests/use_case/asr/Wav2LetterPreprocessingTest.cc +++ b/tests/use_case/asr/Wav2LetterPreprocessingTest.cc @@ -108,7 +108,7 @@ TEST_CASE("Preprocessing calculation INT8") /* Constants. */ const uint32_t windowLen = 512; const uint32_t windowStride = 160; - const int dimArray[] = {3, 1, numMfccFeatures * 3, numMfccVectors}; + int dimArray[] = {3, 1, numMfccFeatures * 3, numMfccVectors}; const float quantScale = 0.1410219967365265; const int quantOffset = -11; diff --git a/tests/use_case/img_class/InferenceTestMobilenetV2.cc b/tests/use_case/img_class/InferenceTestMobilenetV2.cc index b2720a8..6fbf374 100644 --- a/tests/use_case/img_class/InferenceTestMobilenetV2.cc +++ b/tests/use_case/img_class/InferenceTestMobilenetV2.cc @@ -24,7 +24,7 @@ using namespace test; -bool RunInference(arm::app::Model& model, const uint8_t imageData[]) +bool RunInference(arm::app::Model& model, const int8_t imageData[]) { TfLiteTensor* inputTensor = model.GetInputTensor(0); REQUIRE(inputTensor); diff --git a/tests/use_case/kws_asr/InferenceTestWav2Letter.cc b/tests/use_case/kws_asr/InferenceTestWav2Letter.cc index 897ad0a..5f5ad98 100644 --- a/tests/use_case/kws_asr/InferenceTestWav2Letter.cc +++ b/tests/use_case/kws_asr/InferenceTestWav2Letter.cc @@ -55,8 +55,7 @@ bool RunInferenceRandom(arm::app::Model& model) 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][.]") +TEST_CASE("Running random inference with Tflu and Wav2LetterModel Int8", "[Wav2Letter]") { arm::app::Wav2LetterModel model{}; @@ -88,7 +87,7 @@ void TestInference(const T* input_goldenFV, const T* output_goldenFV, arm::app:: } } -TEST_CASE("Running inference with Tflu and Wav2LetterModel Int8", "[Wav2Letter][.]") +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);; diff --git a/tests/use_case/kws_asr/Wav2LetterPreprocessingTest.cc b/tests/use_case/kws_asr/Wav2LetterPreprocessingTest.cc index e71366a..16dbea2 100644 --- a/tests/use_case/kws_asr/Wav2LetterPreprocessingTest.cc +++ b/tests/use_case/kws_asr/Wav2LetterPreprocessingTest.cc @@ -108,7 +108,7 @@ TEST_CASE("Preprocessing calculation INT8") /* Constants. */ const uint32_t windowLen = 512; const uint32_t windowStride = 160; - const int dimArray[] = {3, 1, numMfccFeatures * 3, numMfccVectors}; + int dimArray[] = {3, 1, numMfccFeatures * 3, numMfccVectors}; const float quantScale = 0.1410219967365265; const int quantOffset = -11; -- cgit v1.2.1