三、ONNX Runtime添加一个新的execution provider

ONNX Runtime添加一个新的execution provider

execution provider从本质上来讲就是一个针对不同硬件平台的executor，ONNX Runtime目前提供了以下

• MLAS (Microsoft Linear Algebra Subprograms)

• NVIDIA CUDA

• Intel MKL-ML

• Intel DNNL - subgraph optimization

• Intel nGraph

• NVIDIA TensorRT

• Intel OpenVINO

• Nuphar Model Compiler

• DirectML

• ACL (in preview, for ARM Compute Library)
  10种特定的execution provider，涵盖了ARM，Intel CPU，Nvidia的GPU，不同的操作系统，但是这对于目前纷繁的AI硬件市场远远不够，假如我们自己想添加一个自己定制平台的execution provider，就需要安装如下教程添加自己的execution provider。

1. 在 onnxruntime/core/providers下创建一个新的文件夹，取个名字如your_provider_name。
2. 在 include/onnxruntime/core/providers下创建一个新的文件夹, 名字应该和第一步一样为your_provider_name.
3. 创建一个新的类，必须继承于IExecutionProvider。实现的源码需要放在’onnxruntime/core/providers/[your_provider_name]'目录下。
4. 在 include/onnxruntime/core/providers/[your_provider_name]目录下创建一个头文件。这个有文件应该提供创建OrtProviderFactoryInterface的函数接口。你可以使用 'include/onnxruntime/core/providers/cpu/cpu_provider_factory.h’作为模板，需要注意的是创建 MemoryInfo函数不是必须的。
5. 在 'onnxruntime/core/providers/[your_provider_name]'目录下面创建一个symbols.txt文件，该文件应该包含本execution provider导出的所有的函数名。通常只需一个函数即可创建execution provider工厂。
6. 在onnxruntime_providers.cmake文件中增加你的execution provider编译代码，并编译成一个静态库。
7. 在 cmake/onnxruntime.cmake增加一行target_link_libraries函数，使得onnxruntime能链接上你的execution provider库。

• Create a folder under onnxruntime/core/providers
• Create a folder under include/onnxruntime/core/providers, it should has the same name as the first step.
• Create a new class, which must inherit from IExecutionProvider. The source code should be put in ‘onnxruntime/core/providers/[your_provider_name]’
• Create a new header file under include/onnxruntime/core/providers/[your_provider_name]. The file should provide one function for creating an OrtProviderFactoryInterface. You may use ‘include/onnxruntime/core/providers/cpu/cpu_provider_factory.h’ as a template. You don’t need to provide a function for creating MemoryInfo.
• Put a symbols.txt under ‘onnxruntime/core/providers/[your_provider_name]’. The file should contain all the function names that would be exported from you provider. Usually, just a single function for creating provider factory is enough.
• Add your provider in onnxruntime_providers.cmake. Build it as a static lib.
• Add one line in cmake/onnxruntime.cmake, to the ‘target_link_libraries’ function call. Put your provider there.

Examples:

• CPU Execution
  Provider
• CUDA Execution
  Provider
• DNNL Execution
  Provider

Using the execution provider

1. Create a factory for that provider, by using the c function you exported in ‘symbols.txt’
2. Put the provider factory into session options
3. Create session from that session option
   e.g.

  OrtEnv* env;
  OrtInitialize(ORT_LOGGING_LEVEL_WARNING, "test", &env)
  OrtSessionOptions* session_option = OrtCreateSessionOptions();
  OrtProviderFactoryInterface** factory;
  OrtCreateCUDAExecutionProviderFactory(0, &factory);
  OrtSessionOptionsAppendExecutionProvider(session_option, factory);
  OrtReleaseObject(factory);
  OrtCreateSession(env, model_path, session_option, &session);