microsoft
/
Windows-Machine-Learning
зеркало из
1
0
Форкнуть 0
Код Задачи Пакеты Проекты Релизы Вики Активность

Fix per iteration perf output functionality that was deleted in refactoring work

This commit is contained in:
Ryan Lai 2019-04-17 14:55:33 -07:00
Родитель 31a0ec2c03
Коммит 9ccd65377b
3 изменённых файлов: 54 добавлений и 43 удалений
Показать статистику Скачать Patch файл Скачать Diff файл Показать все файлы Свернуть все файлы

19
Tools/WinMLRunner/src/CommandLineArgs.cpp
Просмотреть файл

@ -218,13 +218,7 @@ CommandLineArgs::CommandLineArgs(const std::vector<std::wstring>& args)
}
catch (...)
{
// set to default path
auto time = std::time(nullptr);
struct tm localTime;
localtime_s(&localTime, &time);
std::wostringstream oss;
oss << std::put_time(&localTime, L"%Y-%m-%d_%H.%M.%S");
SetTensorOutputPath(L"\\PerIterationRun[" + oss.str() + L"]");
// Will Set Default Path after argument checks
}
}
else if (_wcsicmp(args[i].c_str(), L"-version") == 0)
@ -318,6 +312,17 @@ CommandLineArgs::CommandLineArgs(const std::vector<std::wstring>& args)
throw hresult_invalid_argument(msg.c_str());
}
}
// set default path for per iteration / tensor output
if (this->TensorOutputPath().empty())
{
auto time = std::time(nullptr);
struct tm localTime;
localtime_s(&localTime, &time);
std::wostringstream oss;
oss << std::put_time(&localTime, L"%Y-%m-%d_%H.%M.%S");
SetTensorOutputPath(L".\\PerIterationRun[" + oss.str() + L"]");
}
CheckForInvalidArguments();
}

2
Tools/WinMLRunner/src/OutputHelper.h
Просмотреть файл

@ -596,7 +596,7 @@ public:
void SetCSVFileName(const std::wstring& fileName) { m_csvFileName = fileName; }
void WritePerIterationPerformance(const CommandLineArgs& args, std::wstring model, std::wstring img)
void WritePerIterationPerformance(const CommandLineArgs& args, const std::wstring& model, const std::wstring& img)
{
if (m_csvFileNamePerIterationSummary.length() > 0)
{

42
Tools/WinMLRunner/src/Run.cpp
Просмотреть файл

@ -80,8 +80,8 @@ HRESULT BindInputFeatures(const LearningModel& model, const LearningModelBinding
return S_OK;
}
HRESULT LoadModel(LearningModel &model, const std::wstring& path, bool capturePerf, OutputHelper& output, const CommandLineArgs& args,
uint32_t iterationNum, Profiler<WINML_MODEL_TEST_PERF>& profiler)
HRESULT LoadModel(LearningModel& model, const std::wstring& path, bool capturePerf, OutputHelper& output,
const CommandLineArgs& args, uint32_t iterationNum, Profiler<WINML_MODEL_TEST_PERF>& profiler)
{
try
{
@ -111,9 +111,9 @@ HRESULT LoadModel(LearningModel &model, const std::wstring& path, bool capturePe
return S_OK;
}
HRESULT CreateSession(LearningModelSession& session, IDirect3DDevice& winrtDevice, LearningModel& model, CommandLineArgs& args, OutputHelper& output,
DeviceType deviceType, DeviceCreationLocation deviceCreationLocation,
Profiler<WINML_MODEL_TEST_PERF>& profiler)
HRESULT CreateSession(LearningModelSession& session, IDirect3DDevice& winrtDevice, LearningModel& model,
CommandLineArgs& args, OutputHelper& output, DeviceType deviceType,
DeviceCreationLocation deviceCreationLocation, Profiler<WINML_MODEL_TEST_PERF>& profiler)
{
if (model == nullptr)
{
@ -209,12 +209,14 @@ HRESULT CreateSession(LearningModelSession& session, IDirect3DDevice& winrtDevic
return S_OK;
}
HRESULT BindInputs(LearningModelBinding &context, const LearningModel& model, const LearningModelSession& session, OutputHelper& output,
DeviceType deviceType, const CommandLineArgs& args, InputBindingType inputBindingType, InputDataType inputDataType,
const IDirect3DDevice& winrtDevice, DeviceCreationLocation deviceCreationLocation, uint32_t iteration,
HRESULT BindInputs(LearningModelBinding& context, const LearningModel& model, const LearningModelSession& session,
OutputHelper& output, DeviceType deviceType, const CommandLineArgs& args,
InputBindingType inputBindingType, InputDataType inputDataType, const IDirect3DDevice& winrtDevice,
DeviceCreationLocation deviceCreationLocation, uint32_t iteration,
Profiler<WINML_MODEL_TEST_PERF>& profiler)
{
if (deviceType == DeviceType::CPU && inputDataType == InputDataType::Tensor && inputBindingType == InputBindingType::GPU)
if (deviceType == DeviceType::CPU && inputDataType == InputDataType::Tensor &&
inputBindingType == InputBindingType::GPU)
{
std::cout << "Cannot create D3D12 device on client if CPU device type is selected." << std::endl;
return E_INVALIDARG;
@ -238,7 +240,6 @@ HRESULT BindInputs(LearningModelBinding &context, const LearningModel& model, co
return hr.code();
}
HRESULT bindInputResult =
BindInputFeatures(model, context, inputFeatures, args, output, captureIterationPerf, iteration, profiler);
@ -277,14 +278,14 @@ std::vector<std::wstring> GetModelsInDirectory(CommandLineArgs& args, OutputHelp
return modelPaths;
}
HRESULT CheckIfModelAndConfigurationsAreSupported(LearningModel& model, const std::wstring& modelPath,
const DeviceType deviceType,
const std::vector<InputDataType>& inputDataTypes,
const std::vector<DeviceCreationLocation>& deviceCreationLocations)
{
// Does user want image as input binding
bool hasInputBindingImage = std::any_of(inputDataTypes.begin(), inputDataTypes.end(), [](const InputDataType inputDataType){
bool hasInputBindingImage =
std::any_of(inputDataTypes.begin(), inputDataTypes.end(), [](const InputDataType inputDataType) {
return inputDataType == InputDataType::ImageBGR || inputDataType == InputDataType::ImageRGB;
});
@ -345,7 +346,6 @@ HRESULT EvaluateModel(LearningModelEvaluationResult& result, const LearningModel
if (args.IsPerIterationCapture())
{
output.SaveEvalPerformance(profiler, iterationNum);
}
}
}
@ -500,7 +500,8 @@ int run(CommandLineArgs& args, Profiler<WINML_MODEL_TEST_PERF>& profiler) try
{
LearningModel model = nullptr;
LoadModel(model, path, args.IsPerformanceCapture() || args.IsPerIterationCapture(), output, args, 0, profiler);
LoadModel(model, path, args.IsPerformanceCapture() || args.IsPerIterationCapture(), output, args, 0,
profiler);
for (auto deviceType : deviceTypes)
{
lastHr = CheckIfModelAndConfigurationsAreSupported(model, path, deviceType, inputDataTypes,
@ -542,8 +543,9 @@ int run(CommandLineArgs& args, Profiler<WINML_MODEL_TEST_PERF>& profiler) try
{
#if defined(_AMD64_)
// PIX markers only work on AMD64
// If PIX tool was attached then capture already began for the first iteration before session creation.
// This is to begin PIX capture for each iteration after the first iteration.
// If PIX tool was attached then capture already began for the first iteration before
// session creation. This is to begin PIX capture for each iteration after the first
// iteration.
if (i > 0 && output.GetGraphicsAnalysis())
{
output.GetGraphicsAnalysis()->BeginCapture();
@ -558,8 +560,8 @@ int run(CommandLineArgs& args, Profiler<WINML_MODEL_TEST_PERF>& profiler) try
}
LearningModelEvaluationResult result = nullptr;
bool capture_perf = args.IsPerformanceCapture() || args.IsPerIterationCapture();
lastHr = EvaluateModel(result, model, context, session, args, output,
capture_perf, i, profiler);
lastHr = EvaluateModel(result, model, context, session, args, output, capture_perf, i,
profiler);
if (FAILED(lastHr))
{
output.PrintEvaluatingInfo(i + 1, deviceType, inputBindingType, inputDataType,
@ -604,6 +606,10 @@ int run(CommandLineArgs& args, Profiler<WINML_MODEL_TEST_PERF>& profiler) try
inputBindingTypeStringified,
deviceCreationLocationStringified);
}
else if (args.IsPerIterationCapture() && args.IsImageInput())
{
output.WritePerIterationPerformance(args, args.ModelPath(), args.ImagePath());
}
}
}
}