mirror of
https://github.com/saymrwulf/onnxruntime.git
synced 2026-05-17 21:10:43 +00:00
5ce675c3b9
* Register ILearningModelSessionOptionsNate interface * Threading options exposed * Add interrogator for Session options * Add test * Polish test * PR comments * Set intra op threads * Add adapter api to grab intra op threads * Add adapter test for getting intraop num threads * Make ILearningModelSessionNative and update winml api test * Make it required when building engine to set the intraop num threads * Make test more pretty * Change naming of idl function * Revert "Change naming of idl function" This reverts commit c06916aa5bf94e3bf233ed281e508b935fc8638d. * PR comment on naming * Skip the test because it's influenced if it's built with openmp Co-authored-by: Ryan Lai <ryalai96@gamil.com>
61 lines
2.7 KiB
C++
61 lines
2.7 KiB
C++
// Copyright (c) Microsoft Corporation. All rights reserved.
|
|
// Licensed under the MIT License.
|
|
|
|
#pragma once
|
|
|
|
#include "LearningModelSessionOptions.g.h"
|
|
#include <thread>
|
|
namespace WINMLP {
|
|
|
|
struct LearningModelSessionOptions : LearningModelSessionOptionsT<LearningModelSessionOptions, ILearningModelSessionOptionsNative> {
|
|
LearningModelSessionOptions() = default;
|
|
|
|
LearningModelSessionOptions(const LearningModelSessionOptions& options);
|
|
|
|
uint32_t BatchSizeOverride();
|
|
void BatchSizeOverride(uint32_t value);
|
|
|
|
bool CloseModelOnSessionCreation();
|
|
void CloseModelOnSessionCreation(bool value);
|
|
|
|
STDMETHOD(SetIntraOpNumThreadsOverride)
|
|
(uint32_t intraOpNumThreads);
|
|
|
|
uint32_t GetIntraOpNumThreads();
|
|
|
|
private:
|
|
// The batch size override property is used to inform the engine when the developer
|
|
// wants to explicitly set the batch size of a model to a fixed batch size.
|
|
//
|
|
// 0 : dont override the model batch definitions
|
|
// 1...n : override the model with the given batch size
|
|
//
|
|
// This value is a unsigned value, and users are not allowed to override models with a free batch size.
|
|
// If the model supports free dimensional batch sizes, the caller should provide 0, to not override.
|
|
//
|
|
// The default value here is 1 so that models with free dimension batch sizes (which is very common)
|
|
// can be optimized to fixed sizes.
|
|
uint32_t batch_size_override_ = 1;
|
|
|
|
// The close model on session creation property is used to inform the engine when the developer
|
|
// no longer needs the learningmodelsession after session creation.
|
|
// The engine can use the learning model during session creation to move resources rather than make copies.
|
|
//
|
|
// True : Move resources in the LearningModel in to the LearningModelSession
|
|
// False : Copy resources in the LearningModel to the LearningModelSession
|
|
//
|
|
// The default value here is False so that models are not automatically closed on session creation.
|
|
bool close_model_on_session_creation_ = false;
|
|
|
|
// The intra operator num threads property is used to control the number of threads used in the threadpool for intra operator calculations.
|
|
// The default value here is the maximum number of logical cores to ensure that the default behavior of WinML always runs the fastest.
|
|
// WARNING: Setting a number higher than the maximum number of logical cores may result in an inefficient threadpool
|
|
uint32_t intra_op_num_threads_override_ = std::thread::hardware_concurrency();
|
|
};
|
|
|
|
} // namespace WINMLP
|
|
|
|
namespace WINML::factory_implementation {
|
|
struct LearningModelSessionOptions : LearningModelSessionOptionsT<LearningModelSessionOptions, implementation::LearningModelSessionOptions> {
|
|
};
|
|
} // namespace WINML::factory_implementation
|