mirror of
https://github.com/saymrwulf/onnxruntime.git
synced 2026-05-26 22:35:43 +00:00
5350abe19d
this is a big PR. we are going to move it up to layer_dev , which is still a L3 so we are still safe to do work there agile. we are going to move this into the L3 so that ryan can start doing intergration testing. we will pause for a full code review and integration test result prior to going into the L2. >>>> raw comments from previous commits >>> * LearningModelSession is cleaned up to use the adapter, and parts of binding are. * moved everything in the winmladapter made it all nano-com using, WRL to construct objects in the ORT side. base interfaces for everythign for winml to call cleaned up a bunch of winml to use the base interfaces. * more pieces * GetData across the abi. * renamed some namepsace cleaned up OrtValue cleaned up Tensor cleaned up custom ops. everything *but* learnignmodel should be clean * make sure it's building. winml.dll is still a monolith.
91 lines
No EOL
2.7 KiB
C++
91 lines
No EOL
2.7 KiB
C++
// Copyright (c) Microsoft Corporation.
|
|
// Licensed under the MIT License.
|
|
|
|
#include "pch.h"
|
|
|
|
// Needed to work around the fact that OnnxRuntime defines ERROR
|
|
#ifdef ERROR
|
|
#undef ERROR
|
|
#endif
|
|
#include "core/session/inference_session.h"
|
|
// Restore ERROR define
|
|
#define ERROR 0
|
|
|
|
#include "CpuOrtSessionBuilder.h"
|
|
#include "inc/WinMLAdapter.h"
|
|
|
|
// winml includes
|
|
#include "core/providers/dml/GraphTransformers/GraphTransformerHelpers.h"
|
|
|
|
// ort includes
|
|
#include "core/providers/cpu/cpu_execution_provider.h"
|
|
#include "core/optimizer/conv_activation_fusion.h"
|
|
#include "core/optimizer/gemm_activation_fusion.h"
|
|
|
|
using namespace Windows::AI::MachineLearning;
|
|
|
|
namespace Windows::AI::MachineLearning::Adapter {
|
|
|
|
CpuOrtSessionBuilder::CpuOrtSessionBuilder() {
|
|
|
|
}
|
|
|
|
HRESULT
|
|
CpuOrtSessionBuilder::CreateSessionOptions(
|
|
onnxruntime::SessionOptions* p_options) {
|
|
RETURN_HR_IF_NULL(E_POINTER, p_options);
|
|
|
|
*p_options = onnxruntime::SessionOptions();
|
|
p_options->graph_optimization_level = onnxruntime::TransformerLevel::Level3;
|
|
|
|
// Onnxruntime will use half the number of concurrent threads supported on the system
|
|
// by default. This causes MLAS to not exercise every logical core.
|
|
// We force the thread pool size to be maxxed out to ensure that WinML always
|
|
// runs the fastest.
|
|
p_options->intra_op_num_threads = std::thread::hardware_concurrency();
|
|
|
|
return S_OK;
|
|
}
|
|
|
|
HRESULT
|
|
CpuOrtSessionBuilder::CreateSession(
|
|
const onnxruntime::SessionOptions& options,
|
|
_winmla::IInferenceSession** p_session,
|
|
onnxruntime::IExecutionProvider** pp_provider) {
|
|
RETURN_HR_IF_NULL(E_POINTER, p_session);
|
|
RETURN_HR_IF_NULL(E_POINTER, pp_provider);
|
|
RETURN_HR_IF(E_POINTER, *pp_provider != nullptr);
|
|
|
|
// Create the inference session
|
|
auto session = std::make_unique<onnxruntime::InferenceSession>(options);
|
|
|
|
// Create the cpu execution provider
|
|
onnxruntime::CPUExecutionProviderInfo xpInfo;
|
|
#ifndef _WIN64
|
|
xpInfo.create_arena = false;
|
|
#endif
|
|
auto cpu_provider = std::make_unique<onnxruntime::CPUExecutionProvider>(xpInfo);
|
|
|
|
// Cache the provider's raw pointer
|
|
*pp_provider = cpu_provider.get();
|
|
|
|
// Register the cpu xp
|
|
ORT_THROW_IF_ERROR(session->RegisterExecutionProvider(std::move(cpu_provider)));
|
|
|
|
// assign the session to the out parameter
|
|
auto sessionptr = wil::MakeOrThrow<_winmla::InferenceSession>(session.release());
|
|
RETURN_IF_FAILED(sessionptr.CopyTo(_uuidof(_winmla::IInferenceSession), (void**)p_session));
|
|
|
|
return S_OK;
|
|
}
|
|
|
|
HRESULT
|
|
CpuOrtSessionBuilder::Initialize(
|
|
_winmla::IInferenceSession* p_session,
|
|
onnxruntime::IExecutionProvider* /*p_provider*/
|
|
) {
|
|
ORT_THROW_IF_ERROR(p_session->get()->Initialize());
|
|
return S_OK;
|
|
}
|
|
|
|
} // Windows::AI::MachineLearning::Adapter
|