[GNA] Moved PWL functional tests (#10731)
* Moving PWL to ngraph * improving the running time of php_search; refactoring the pwl operation * fixed erros & refactored code * moved PWL op to GNA * Update src/plugins/intel_gna/ops/pwl.hpp Co-authored-by: Elizaveta Lobanova <elizaveta.lobanova@intel.com> * Update src/plugins/intel_gna/ops/reference/pwl.hpp Co-authored-by: Elizaveta Lobanova <elizaveta.lobanova@intel.com> * Update src/plugins/intel_gna/ops/pwl.cpp Co-authored-by: Elizaveta Lobanova <elizaveta.lobanova@intel.com> * Update src/plugins/intel_gna/transformations/transpose_to_pwl.hpp Co-authored-by: Elizaveta Lobanova <elizaveta.lobanova@intel.com> * Update src/plugins/intel_gna/transformations/transpose_to_pwl.cpp Co-authored-by: Elizaveta Lobanova <elizaveta.lobanova@intel.com> * fixed compilation error * Update inference-engine/tests/unit/gna/ngraph/transformations/gna_pwl.cpp Co-authored-by: Elizaveta Lobanova <elizaveta.lobanova@intel.com> * added some tests; changed algorithm of checking accuracy of pwl; refactoring * added first and last segments; added fq and fixed errors * fixed after review & rewrote some tests on ngraph * removed debug logs & fixed code style check error * s/ngraph_helper/ngraph_util * removed TRANSFORMATIONS_API in PWLApproximation class declaration * removed OPENVINO_API in Pwl class declaration * replaced the deprecated version of evaluate() with a new one * fixed some problems after reviewing * fixed a problem when a value of function of left point of segment is less than minimum of function * corrected a value of the right point of last segments * [GNA] Moved pwl func tests * Deleted deprecated test * s/OPENVINO_RTTI/OPENVINO_OP * Deleted conflicted test file * fixed after review Co-authored-by: Dmitrii Khurtin <dmitrii.khurtin@intel.com> Co-authored-by: Elizaveta Lobanova <elizaveta.lobanova@intel.com>
This commit is contained in:
Andrey Noskov
GitHub
@@ -95,7 +95,8 @@ const std::map<ActivationTypes, std::vector<std::vector<float>>> activationTypes
|
||||
{Exp, {}},
|
||||
{Log, {}},
|
||||
{Sign, {}},
|
||||
{Abs, {}}
|
||||
{Abs, {}},
|
||||
{Clamp, {{-5, 5}}}
|
||||
};
|
||||
|
||||
const std::map<ActivationTypes, std::vector<std::vector<float>>> preluActivationTypes = {
|
||||
|
||||
@@ -1,248 +0,0 @@
|
||||
// Copyright (C) 2018-2022 Intel Corporation
|
||||
// SPDX-License-Identifier: Apache-2.0
|
||||
//
|
||||
|
||||
#include <vector>
|
||||
#include <gtest/gtest.h>
|
||||
#include "gna_matcher.hpp"
|
||||
#include <ngraph/function.hpp>
|
||||
#include <ngraph/opsets/opset8.hpp>
|
||||
|
||||
namespace detail {
|
||||
template<typename T>
|
||||
std::shared_ptr<ngraph::Function> CreateActivationFunction(const ngraph::Shape& input_shape) {
|
||||
auto input_params = std::make_shared<ngraph::opset8::Parameter>(ngraph::element::f32, input_shape);
|
||||
auto f = std::make_shared<T>(input_params);
|
||||
auto result = std::make_shared<ngraph::opset8::Result>(f);
|
||||
return std::make_shared<ngraph::Function>(ngraph::ResultVector{result}, ngraph::ParameterVector{input_params});
|
||||
}
|
||||
|
||||
template<typename T>
|
||||
std::shared_ptr<ngraph::Function> CreateActivationFunction(const ngraph::Shape& input_shape, double min, double max);
|
||||
|
||||
template<>
|
||||
std::shared_ptr<ngraph::Function> CreateActivationFunction<ngraph::opset8::Clamp>(const ngraph::Shape& input_shape, double min, double max) {
|
||||
auto input_params = std::make_shared<ngraph::opset8::Parameter>(ngraph::element::f32, input_shape);
|
||||
auto f = std::make_shared<ngraph::opset8::Clamp>(input_params, min, max);
|
||||
auto result = std::make_shared<ngraph::opset8::Result>(f);
|
||||
return std::make_shared<ngraph::Function>(ngraph::ResultVector{result}, ngraph::ParameterVector{input_params});
|
||||
}
|
||||
}
|
||||
|
||||
class PWLAproximationTest : public GNATest<> {
|
||||
protected:
|
||||
void SetUp() override {
|
||||
}
|
||||
};
|
||||
using namespace GNATestIRs;
|
||||
|
||||
// Recursive Algorithm
|
||||
// Precision Threshold
|
||||
|
||||
TEST_F(PWLAproximationTest, forTanhOnRecursiveAlgoWithPrecisionThresholdIsSuccess) {
|
||||
assert_that().onInferNgraphModel(detail::CreateActivationFunction<ngraph::opset8::Tanh>({1, 10}))
|
||||
.gna()
|
||||
.withGNAConfig(GNA_CONFIG_KEY(SCALE_FACTOR), 1.0f)
|
||||
.propagate_forward()
|
||||
.called_with()
|
||||
.pwl_quantization_activation(DnnActivationType::kActTanh)
|
||||
.pwl_quantization_precision_threshold(0.0053);
|
||||
}
|
||||
|
||||
TEST_F(PWLAproximationTest, forSigmoidOnRecursiveAlgoWithPrecisionThresholdIsSuccess) {
|
||||
assert_that().onInferNgraphModel(detail::CreateActivationFunction<ngraph::opset8::Sigmoid>({1, 10}))
|
||||
.gna()
|
||||
.withGNAConfig(GNA_CONFIG_KEY(SCALE_FACTOR), 1.0f)
|
||||
.propagate_forward()
|
||||
.called_with()
|
||||
.pwl_quantization_activation(DnnActivationType::kActSigmoid)
|
||||
.pwl_quantization_precision_threshold(0.0027);
|
||||
}
|
||||
|
||||
TEST_F(PWLAproximationTest, forReLUonRecursiveAlgoWithPrecisionThresholdIsSuccess) {
|
||||
assert_that().onInferNgraphModel(detail::CreateActivationFunction<ngraph::opset8::Relu>({1, 1, 10, 10}))
|
||||
.gna()
|
||||
.withGNAConfig(GNA_CONFIG_KEY(SCALE_FACTOR), 1.0f)
|
||||
.propagate_forward()
|
||||
.called_with()
|
||||
.pwl_quantization_activation(DnnActivationType::kActRelu)
|
||||
.pwl_quantization_precision_threshold(0.0001);
|
||||
}
|
||||
|
||||
TEST_F(PWLAproximationTest, forLeakyReLUonRecursiveAlgoWithPrecisionThresholdIsSuccess) {
|
||||
assert_that().onInferModel(LeakyReLUActivationModel())
|
||||
.gna()
|
||||
.withGNAConfig(GNA_CONFIG_KEY(SCALE_FACTOR), 1.0f)
|
||||
.propagate_forward()
|
||||
.called_with()
|
||||
.pwl_quantization_activation(DnnActivationType::kActLeakyRelu)
|
||||
.pwl_quantization_precision_threshold(0.0003);
|
||||
}
|
||||
|
||||
TEST_F(PWLAproximationTest, DISABLED_forIdentityOnRecursiveAlgoWithPrecisionThresholdIsSuccess) {
|
||||
assert_that().onInferModel(IdentityActivationModel())
|
||||
.gna()
|
||||
.withGNAConfig(GNA_CONFIG_KEY(SCALE_FACTOR), 1.0f)
|
||||
.propagate_forward()
|
||||
.called_with()
|
||||
.pwl_quantization_activation(DnnActivationType::kActIdentity)
|
||||
.pwl_quantization_precision_threshold(0.0003);
|
||||
}
|
||||
|
||||
TEST_F(PWLAproximationTest, forClampOnRecursiveAlgoWithPrecisionThresholdIsSuccess) {
|
||||
assert_that().onInferNgraphModel(detail::CreateActivationFunction<ngraph::opset8::Clamp>({1, 10}, -50, 50))
|
||||
.gna()
|
||||
.withGNAConfig(GNA_CONFIG_KEY(SCALE_FACTOR), 1.0f)
|
||||
.propagate_forward()
|
||||
.called_with()
|
||||
.pwl_quantization_activation(DnnActivationType::kActKaldiLstmClipping)
|
||||
.pwl_quantization_precision_threshold(0.0001);
|
||||
}
|
||||
|
||||
TEST_F(PWLAproximationTest, forTanhOnUniformAlgoWithPrecisionThresholdIsSuccess) {
|
||||
assert_that().onInferNgraphModel(detail::CreateActivationFunction<ngraph::opset8::Tanh>({1, 10}))
|
||||
.gna()
|
||||
.withGNAConfig(GNA_CONFIG_KEY(SCALE_FACTOR), 1.0f)
|
||||
.withUniformPWLAlgo()
|
||||
.propagate_forward()
|
||||
.called_with()
|
||||
.pwl_quantization_activation(DnnActivationType::kActTanh)
|
||||
.pwl_quantization_precision_threshold(0.0009);
|
||||
}
|
||||
|
||||
TEST_F(PWLAproximationTest, forSigmoidOnUniformAlgoWithPrecisionThresholdIsSuccess) {
|
||||
assert_that().onInferNgraphModel(detail::CreateActivationFunction<ngraph::opset8::Sigmoid>({1, 10}))
|
||||
.gna()
|
||||
.withGNAConfig(GNA_CONFIG_KEY(SCALE_FACTOR), 1.0f)
|
||||
.withUniformPWLAlgo()
|
||||
.propagate_forward()
|
||||
.called_with()
|
||||
.pwl_quantization_activation(DnnActivationType::kActSigmoid)
|
||||
.pwl_quantization_precision_threshold(0.0004);
|
||||
}
|
||||
|
||||
TEST_F(PWLAproximationTest, DISABLED_forIdentityOnUniformAlgoWithPrecisionThresholdIsSuccess) {
|
||||
assert_that().onInferModel(IdentityActivationModel())
|
||||
.gna()
|
||||
.withGNAConfig(GNA_CONFIG_KEY(SCALE_FACTOR), 1.0f)
|
||||
.withUniformPWLAlgo()
|
||||
.propagate_forward()
|
||||
.called_with()
|
||||
.pwl_quantization_activation(DnnActivationType::kActIdentity)
|
||||
.pwl_quantization_precision_threshold(0.0003);
|
||||
}
|
||||
|
||||
TEST_F(PWLAproximationTest, forClampOnUniformAlgoWithPrecisionThresholdIsSuccess) {
|
||||
assert_that().onInferNgraphModel(detail::CreateActivationFunction<ngraph::opset8::Clamp>({1, 10}, -50, 50))
|
||||
.gna()
|
||||
.withGNAConfig(GNA_CONFIG_KEY(SCALE_FACTOR), 1.0f)
|
||||
.withUniformPWLAlgo()
|
||||
.propagate_forward()
|
||||
.called_with()
|
||||
.pwl_quantization_activation(DnnActivationType::kActKaldiLstmClipping)
|
||||
.pwl_quantization_precision_threshold(0.0001);
|
||||
}
|
||||
|
||||
TEST_F(PWLAproximationTest, forSigmoidonRecursiveAlgoWithSegmentThresholdIsSuccess) {
|
||||
assert_that().onInferNgraphModel(detail::CreateActivationFunction<ngraph::opset8::Sigmoid>({1, 10}))
|
||||
.gna()
|
||||
.withGNAConfig(GNA_CONFIG_KEY(SCALE_FACTOR), 1.0f)
|
||||
.propagate_forward()
|
||||
.called_with()
|
||||
.pwl_quantization_activation(DnnActivationType::kActSigmoid)
|
||||
.pwl_quantization_segments_threshold(12);
|
||||
}
|
||||
|
||||
TEST_F(PWLAproximationTest, forTanhonRecursiveAlgoWithSegmentThresholdIsSuccess) {
|
||||
assert_that().onInferNgraphModel(detail::CreateActivationFunction<ngraph::opset8::Tanh>({1, 10}))
|
||||
.gna()
|
||||
.withGNAConfig(GNA_CONFIG_KEY(SCALE_FACTOR), 1.0f)
|
||||
.propagate_forward()
|
||||
.called_with()
|
||||
.pwl_quantization_activation(DnnActivationType::kActTanh)
|
||||
.pwl_quantization_segments_threshold(12);
|
||||
}
|
||||
|
||||
TEST_F(PWLAproximationTest, forReLUonRecursiveAlgoWithSegmentThresholdIsSuccess) {
|
||||
assert_that().onInferNgraphModel(detail::CreateActivationFunction<ngraph::opset8::Relu>({1, 1, 10, 10}))
|
||||
.gna()
|
||||
.withGNAConfig(GNA_CONFIG_KEY(SCALE_FACTOR), 1.0f)
|
||||
.propagate_forward()
|
||||
.called_with()
|
||||
.pwl_quantization_activation(DnnActivationType::kActRelu)
|
||||
.pwl_quantization_segments_threshold(4);
|
||||
}
|
||||
|
||||
TEST_F(PWLAproximationTest, forLeakyReLUonRecursiveAlgoWithSegmentThresholdIsSuccess) {
|
||||
assert_that().onInferModel(LeakyReLUActivationModel())
|
||||
.gna()
|
||||
.withGNAConfig(GNA_CONFIG_KEY(SCALE_FACTOR), 1.0f)
|
||||
.propagate_forward()
|
||||
.called_with()
|
||||
.pwl_quantization_activation(DnnActivationType::kActLeakyRelu)
|
||||
.pwl_quantization_segments_threshold(4);
|
||||
}
|
||||
|
||||
TEST_F(PWLAproximationTest, DISABLED_forIdentityOnRecursiveAlgoWithSegmentThresholdIsSuccess) {
|
||||
assert_that().onInferModel(IdentityActivationModel())
|
||||
.gna()
|
||||
.withGNAConfig(GNA_CONFIG_KEY(SCALE_FACTOR), 1.0f)
|
||||
.propagate_forward()
|
||||
.called_with()
|
||||
.pwl_quantization_activation(DnnActivationType::kActIdentity)
|
||||
.pwl_quantization_segments_threshold(3);
|
||||
}
|
||||
|
||||
TEST_F(PWLAproximationTest, forClampOnRecursiveAlgoWithSegmentThresholdIsSuccess) {
|
||||
assert_that().onInferNgraphModel(detail::CreateActivationFunction<ngraph::opset8::Clamp>({1, 10}, -50, 50))
|
||||
.gna()
|
||||
.withGNAConfig(GNA_CONFIG_KEY(SCALE_FACTOR), 1.0f)
|
||||
.propagate_forward()
|
||||
.called_with()
|
||||
.pwl_quantization_activation(DnnActivationType::kActKaldiLstmClipping)
|
||||
.pwl_quantization_segments_threshold(3);
|
||||
}
|
||||
|
||||
TEST_F(PWLAproximationTest, forSigmoidonUniformAlgoWithSegmentThresholdIsSuccess) {
|
||||
assert_that().onInferNgraphModel(detail::CreateActivationFunction<ngraph::opset8::Sigmoid>({1, 10}))
|
||||
.gna()
|
||||
.withGNAConfig(GNA_CONFIG_KEY(SCALE_FACTOR), 1.0f)
|
||||
.withUniformPWLAlgo()
|
||||
.propagate_forward()
|
||||
.called_with()
|
||||
.pwl_quantization_activation(DnnActivationType::kActSigmoid)
|
||||
.pwl_quantization_segments_threshold(65);
|
||||
}
|
||||
|
||||
TEST_F(PWLAproximationTest, forTanhonUniformAlgoWithSegmentThresholdIsSuccess) {
|
||||
assert_that().onInferNgraphModel(detail::CreateActivationFunction<ngraph::opset8::Tanh>({1, 10}))
|
||||
.gna()
|
||||
.withGNAConfig(GNA_CONFIG_KEY(SCALE_FACTOR), 1.0f)
|
||||
.withUniformPWLAlgo()
|
||||
.propagate_forward()
|
||||
.called_with()
|
||||
.pwl_quantization_activation(DnnActivationType::kActTanh)
|
||||
.pwl_quantization_segments_threshold(65);
|
||||
}
|
||||
|
||||
TEST_F(PWLAproximationTest, DISABLED_forIdentityOnUniformAlgoWithSegmentThresholdIsSuccess) {
|
||||
assert_that().onInferModel(IdentityActivationModel())
|
||||
.gna()
|
||||
.withGNAConfig(GNA_CONFIG_KEY(SCALE_FACTOR), 1.0f)
|
||||
.withUniformPWLAlgo()
|
||||
.propagate_forward()
|
||||
.called_with()
|
||||
.pwl_quantization_activation(DnnActivationType::kActIdentity)
|
||||
.pwl_quantization_segments_threshold(3);
|
||||
}
|
||||
|
||||
TEST_F(PWLAproximationTest, forClampOnUniformAlgoWithSegmentThresholdIsSuccess) {
|
||||
assert_that().onInferNgraphModel(detail::CreateActivationFunction<ngraph::opset8::Clamp>({1, 10}, -50, 50))
|
||||
.gna()
|
||||
.withGNAConfig(GNA_CONFIG_KEY(SCALE_FACTOR), 1.0f)
|
||||
.withUniformPWLAlgo()
|
||||
.propagate_forward()
|
||||
.called_with()
|
||||
.pwl_quantization_activation(DnnActivationType::kActKaldiLstmClipping)
|
||||
.pwl_quantization_segments_threshold(3);
|
||||
}
|
||||
Reference in New Issue
Block a user