Python模型转为C++模型 (PyTorch )

PyTorch模型主要用Python语言开发，在工业界，往往需要将Python的模型转化为C++，以方便其部署和跨平台使用；

1.创建PyTorch模型

2.转换成torch_script

3.保存pt模型

4.C++Load模型

5.编译

编译cmake的版本问题，下载和更新cmake的版本将会是比较麻烦、甚至看运气的事情，根据当前的torch版本，和CPU/GPU环境，选择libtorch版本

编辑CMakeList.txt文件

cmake_minimum_required(VERSION 3.0 FATAL_ERROR)
project(example_app)

find_package(Torch REQUIRED)

add_executable(example-app example-app.cpp)
target_link_libraries(example-app "${TORCH_LIBRARIES}")
set_property(TARGET example-app PROPERTY CXX_STANDARD 14)

$ cmake -DCMAKE_PREFIX_PATH=/Users/xxx/CLionProjects/libtorch ..

-- The C compiler identification is AppleClang 9.1.0.9020039

-- The CXX compiler identification is AppleClang 9.1.0.9020039

-- Detecting C compiler ABI info

-- Detecting C compiler ABI info - done

-- Check for working C compiler: /Library/Developer/CommandLineTools/usr/bin/cc - skipped

-- Detecting C compile features

-- Detecting C compile features - done

-- Detecting CXX compiler ABI info

-- Detecting CXX compiler ABI info - done

-- Check for working CXX compiler: /Library/Developer/CommandLineTools/usr/bin/c++ - skipped

-- Detecting CXX compile features

-- Detecting CXX compile features - done

-- Looking for pthread.h

-- Looking for pthread.h - found

-- Performing Test CMAKE_HAVE_LIBC_PTHREAD

-- Performing Test CMAKE_HAVE_LIBC_PTHREAD - Success

-- Found Threads: TRUE  

-- Found torch: /Users/tiger/CLionProjects/libtorch/lib/libtorch.dylib  

-- Configuring done

-- Generating done

-- Build files have been written to: /Users/xxx/CLionProjects/example-app/build

$ cmake --build . --config Release

Scanning dependencies of target example-app

[ 50%] Building CXX object CMakeFiles/example-app.dir/example-app.cpp.o

[100%] Linking CXX executable example-app

[100%] Built target example-app

6.测试效果

./example-app /Users/xxx/PycharmProjects/ReadCamera/model1.pt 

ok

-0.0913 -0.2469  0.3947 -0.0506  0.0468

7.模型在GPU里运行

model.to(at::kCUDA)

tensor.to(at::kCUDA)