厌倦了Python训练？C++版本PyTorch尝试一下

2023-10-18 23:45| 来源: 网络整理| 查看: 265

喜欢就关注我们吧！

PyTorch的流行给深度学习的发展带来了巨大的便利，然而工业界和嵌入式平台对于深度学习的应用还远远不够，很重要的一个原因就是Python在效率方面不能满足工业界的需求。PyTorch 1.5正式版的发布则是填补了PyTorch在这个方面的空缺。今天的一篇文章就为大家简单介绍一下如何用C++版本的PyTorch来训练。

原文作者：Venkata Chintapalli

翻译：zdi

PyTorch 1.5稳定版的一个重要更新就是增加了对于C++前端API的支持。C++API的支持对于要求低延迟，多线程，高并发的系统的很有必要。PyTroch的文档中对此有详细的说明（点击阅读原文即可查看）。本文以一个两层的神经网络为例带大家入门PyTorch的C++用法，后续还会为大家带来更多有关PyTorch的C++使用的教程。

Libtorch C++库的安装

首先介绍一下如何安装，安装其实很简单，只需要从PyTorch官方下载代码，解压编译即可。

wget https://download.pytorch.org/libtorch/nightly/cpu/libtorch-shared-with-deps-latest.zip unzip libtorch-shared-with-deps-latest.zip

这里只是针对最新版本的配置，对于不同的硬件条件和深度学习环境，请大家参考https://pytorch.org/tutorials/advanced/cpp_frontend.html查看。

下载完成之后的步骤是编译libtorch。

//build the example mkdir build cmake -DCMAKE_PREFIX_PATH=/path/to/libtorch // run the example cmake --build . --config Release

正常的输出为：

$ cmake -DCMAKE_PREFIX_PATH=~/pytorch/libtorch CMake Warning: No source or binary directory provided. Both will be assumed to be the same as the current working directory, but note that this warning will become a fatal error in future CMake releases. -- The C compiler identification is GNU 7.5.0 -- The CXX compiler identification is GNU 7.5.0 -- Check for working C compiler: /usr/bin/cc -- Check for working C compiler: /usr/bin/cc -- works -- Detecting C compiler ABI info -- Detecting C compiler ABI info - done -- Detecting C compile features -- Detecting C compile features - done -- Check for working CXX compiler: /usr/bin/c++ -- Check for working CXX compiler: /usr/bin/c++ -- works -- Detecting CXX compiler ABI info -- Detecting CXX compiler ABI info - done -- Detecting CXX compile features -- Detecting CXX compile features - done -- Looking for pthread.h -- Looking for pthread.h - found -- Looking for pthread_create -- Looking for pthread_create - not found -- Looking for pthread_create in pthreads -- Looking for pthread_create in pthreads - not found -- Looking for pthread_create in pthread -- Looking for pthread_create in pthread - found -- Found Threads: TRUE -- Found Torch: /home/yasasri/pytorch/libtorch/lib/libtorch.so -- Configuring done -- Generating done -- Build files have been written to: /home/yasasri/code/PyTorchCppFrontEnd (base) yasasri@yasasri-MacBookPro:~/code/PyTorchCppFrontEnd$ cmake --build . --config Release Scanning dependencies of target TwoLayerNetwork [ 50%] Building CXX object CMakeFiles/TwoLayerNetwork.dir/TwoLayerNetwork.cpp.o [100%] Linking CXX executable TwoLayerNetwork [100%] Built target TwoLayerNetwork

张量

PyTorch中最基础的三个模块为：张量，自动求导和网络模块。由于Python是动态语言，因此在Python中声明张量不需要指定变量类型，而C++是一门静态语言，在编码过程中指定变量类型，因此可以更快更安全的执行。因此将PyTorch的Python代码转化为C++代码的第一步就是指定变量类型。现代C++同样支持了 auto 类型的变量类型，编译器会在编译过程中自动为变量分配类型。

假设我们想要构造一个输入输出为随机值的函数

在PyTorch的Python接口中我们的写法为：

# Pythonx = torch.randn(N, D_in)y = torch.randn(N, D_out)

对应的C++写法为：

// C++ torch::Tensor x = torch::rand({N, D_in}); torch::Tensor y = torch::rand({N, D_out});

在这里我们用作用域运算符::取代了Python中.的用法。

两层的神经网络模型

使用Python在PyTorch中定义模型时，需要继承torch.nn.Module。子类会相应继承父类的方法和属性，在Python中定义一个两层的神经网络的代码为：

# Pythonimport torch class TwoLayerNet(torch.nn.Module): def __init__(self, D_in, H, D_out): super(TwoLayerNet, self).__init__() self.linear1 = torch.nn.Linear(D_in, H) self.linear2 = torch.nn.Linear(H, D_out) def forward(self.x): h_relu = self.linear1(x).clamp(min=0) y_pred = self.linear2(h_relu) return y_pred

C++定义神经网络的方法是使用结构体（struct），值语义（value semantics）和引用语义（reference semantics）的都可以用于创建模型。使用值语义的代码为：

// C++ value semantics#include // C++ neural network model defined with value semanticsstruct TwoLayerNet : torch::nn::Module { // constructor with submodules registered in the initializer list TwoLayerNet(int64_t D_in, int64_t D_out, int64_t H) : linear1(register_module("linear1", torch::nn::Linear(D_in, H))), linear2(register_module("linear2", torch::nn::Linear(H, D_out))) {} torch::Tensor forward(torch::Tensor x) { x = torch::relu(linear1->forward(x)); x = linear2->forward(x); return x; } torch::nn::Linear linear1; torch::nn::Linear linear2;};// Usage: access the object using dot . operatorTwoLayerNet model(D_in, D_out, H);model.to(device);model.forward(x);model.parameters();

子类使用register_module() 方法可以调用父类的属性和方法：

/// Registers a submodule with this `Module`.// This method deals with `ModuleHolder`s.// Registering a module makes it available to methods such as /// `modules()`, `clone()` or `to()`.// \rst/// .. code-block:: cpp// MyModule::MyModule() {/// submodule_ = register_module("linear", torch::nn::Linear(3, 4));/// }/// \endrsttemplate std::shared_ptr register_module( std::string name, ModuleHolder module_holder);

使用引用语义定义在C++中定义网络的方法是使用std::shared_ptr。TORCH_MODULE(TwoLayerNet)的宏定义定义了两层的网络结构。PyTorch推荐使用引用语义的方式在C++中定义模型，具体的代码为：

// C++ neural network model defined with reference semanticsstruct TwoLayerNetImpl : torch::nn::Module { // module holders are assigned in the constructor TwoLayerNetImpl(int64_t D_in, int64_t D_out, int64_t H) : linear1(D_in, H), linear2(H, D_out) { register_module("linear1", linear1); register_module("linear2", linear2); } ... // forward() method torch::nn::Linear linear1{nullptr}; //construct an empty holder torch::nn::Linear linear2{nullptr}; //construct an empty holder};TORCH_MODULE(TwoLayerNet);// Usage: access the object using arrow -> operatorTwoLayerNet model(D_in, D_out, H);model->to(device);model->forward(x);model->parameters();

优化器

Python和C++的优化器代码类似：

# Python SGD optimizer learning_rate = 1e-4 optimizer = torch.optim.SGD(model.parameters(), lr=learning_rate, momentum=0.5) # Zero the gradients before running the backward pass optimizer.zero_grad() # Update weights optimizer.step()

C++代码为：

// C++ SGD optimizerfloat_t learning_rate = 1e-4;torch::optim::SGD optimizer( model.parameters(), torch::optim::SGDOptions(learning_rate).momentum(0.5));// Zero the gradients before running the backward passoptimizer.zero_grad()// Update the weightsoptimizer.step()

结果对比

在Ubuntu 18.04的机器上，C++的效率大致是Python的两倍，下图为对比结果