IntenseWebs/openvino
3
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Remove old dynamism

Browse Source
This commit is contained in:
Efode, Irina 2021-10-14 01:34:14 +03:00
parent fadf43ecb7
commit 1393c9c805
2 changed files with 5 additions and 69 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

7
inference-engine/tests/functional/shared_test_classes/include/shared_test_classes/base/layer_test_utils.hpp
View File

@ -155,12 +155,6 @@ protected:
float abs_threshold; float abs_threshold;
InferenceEngine::CNNNetwork cnnNetwork; InferenceEngine::CNNNetwork cnnNetwork;
std::shared_ptr<InferenceEngine::Core> core; std::shared_ptr<InferenceEngine::Core> core;
// dynamic input shapes
std::vector<ngraph::PartialShape> inputDynamicShapes;
// index for targetStaticShape
size_t index = 0;
// target static input shapes which is used for reshape ngraph function & generate input blobs
std::vector<std::vector<ngraph::Shape>> targetStaticShapes;
virtual void Validate(); virtual void Validate();
@ -171,7 +165,6 @@ protected:
InferenceEngine::InferRequest inferRequest; InferenceEngine::InferRequest inferRequest;
private: private:
void ResizeNgraphFunction();
RefMode refMode = RefMode::INTERPRETER; RefMode refMode = RefMode::INTERPRETER;
}; };

67
inference-engine/tests/functional/shared_test_classes/src/base/layer_test_utils.cpp
View File

@ -24,16 +24,6 @@ namespace LayerTestsUtils {
LayerTestsCommon::LayerTestsCommon() : threshold(1e-2f), abs_threshold(-1.f) { LayerTestsCommon::LayerTestsCommon() : threshold(1e-2f), abs_threshold(-1.f) {
core = PluginCache::get().ie(targetDevice); core = PluginCache::get().ie(targetDevice);
} }
void LayerTestsCommon::ResizeNgraphFunction() {
auto params = function->get_parameters();
std::map<std::string, ngraph::PartialShape> shapes;
ASSERT_LE(params.size(), targetStaticShapes[index].size());
for (size_t i = 0; i < params.size(); i++) {
shapes.insert({*params[i]->get_output_tensor(0).get_names().begin(), targetStaticShapes[index][i]});
}
function->reshape(shapes);
functionRefs->reshape(shapes);
}
void LayerTestsCommon::Run() { void LayerTestsCommon::Run() {
if (functionRefs == nullptr) { if (functionRefs == nullptr) {
@ -60,27 +50,10 @@ void LayerTestsCommon::Run() {
try { try {
LoadNetwork(); LoadNetwork();
size_t i = 0; GenerateInputs();
do { Infer();
index = i; Validate();
try { s.updateOPsStats(functionRefs, PassRate::Statuses::PASSED);
if (!inputDynamicShapes.empty()) {
// resize ngraph function according new target shape
ResizeNgraphFunction();
}
GenerateInputs();
Infer();
Validate();
s.updateOPsStats(functionRefs, PassRate::Statuses::PASSED);
} catch (const std::exception &ex) {
std::string errorMessage;
if (!targetStaticShapes.empty()) {
errorMessage = "Incorrect target static shape: " + CommonTestUtils::vec2str(targetStaticShapes[i]) + "\n";
}
errorMessage += ex.what();
THROW_IE_EXCEPTION << ex.what();
}
} while (++i < targetStaticShapes.size());
} }
catch (const std::runtime_error &re) { catch (const std::runtime_error &re) {
s.updateOPsStats(functionRefs, PassRate::Statuses::FAILED); s.updateOPsStats(functionRefs, PassRate::Statuses::FAILED);
@ -364,22 +337,6 @@ void LayerTestsCommon::ConfigureNetwork() {
out.second->setPrecision(outPrc); out.second->setPrecision(outPrc);
} }
} }
// Reshape CNNNetwork before load to the plugin in dynamic scenario
if (!inputDynamicShapes.empty()) {
auto params = function->get_parameters();
std::map<std::string, ngraph::PartialShape> inputDataMap;
ASSERT_EQ(params.size(), inputDynamicShapes.size());
for (size_t i = 0; i < inputDynamicShapes.size(); i++) {
ngraph::PartialShape dynamicShape = inputDynamicShapes[i];
if (dynamicShape.rank() == 0 && dynamicShape.is_static()) {
continue;
}
std::string inputName = params[i]->get_friendly_name();
inputDataMap.insert({inputName, dynamicShape});
}
cnnNetwork.reshape(inputDataMap);
}
} }
void LayerTestsCommon::LoadNetwork() { void LayerTestsCommon::LoadNetwork() {
@ -398,21 +355,7 @@ void LayerTestsCommon::GenerateInputs() {
const auto infoIt = inputsInfo.find(param->get_friendly_name()); const auto infoIt = inputsInfo.find(param->get_friendly_name());
GTEST_ASSERT_NE(infoIt, inputsInfo.cend()); GTEST_ASSERT_NE(infoIt, inputsInfo.cend());
InferenceEngine::InputInfo::CPtr info = infoIt->second; InferenceEngine::InputInfo::CPtr info = infoIt->second;
InferenceEngine::Blob::Ptr blob = nullptr; InferenceEngine::Blob::Ptr blob = GenerateInput(*info);
if (!inputDynamicShapes.empty()) {
if (inputDynamicShapes[i].rank() != 0) {
InferenceEngine::DataPtr dataNew(
new InferenceEngine::Data(infoIt->first, info->getTensorDesc().getPrecision(),
targetStaticShapes[index][i],
info->getTensorDesc().getLayout()));
InferenceEngine::InputInfo infoNew;
infoNew.setInputData(dataNew);
blob = GenerateInput(infoNew);
}
}
if (blob == nullptr) {
blob = GenerateInput(*info);
}
inputs.push_back(blob); inputs.push_back(blob);
} }
} }