首先,下载Libtorch:https://pytorch.org/get-started/locally/
然后打开你的项目,使用CMake构建哦w
例如我的:
cmake_minimum_required(VERSION 3.15) project(TorchFly)
set(CMAKE_CXX_STANDARD 14) set(Torch_DIR ~/SDK/libtorch/share/cmake/Torch)
find_package(Torch REQUIRED)
add_executable(TorchFly src/main.cpp) target_link_libraries(TorchFly ${TORCH_LIBRARIES})
|
其中 set(Torch_DIR ~/SDK/libtorch/share/cmake/Torch) 就是你的Torch的路径,绝对路径与相对路径都可以。add_executable(TorchFly src/main.cpp) 是什么不多说了(((
然后是CPP文件:
#include <torch/torch.h> #include <iostream>
int main() { torch::Tensor tensor = torch::eye(4); std::cout << tensor << std::endl; }
|
好了,运行就会输出:
1 0 0 0 0 1 0 0 0 0 1 0 0 0 0 1 [ Variable[CPUFloatType]{4,4} ]
|
如果出现:
dyld: Library not loaded: @rpath/libmklml.dylib Referenced from: ~/libtorch/lib/libtorch.dylib Reason: image not found
|
下载:https://github.com/intel/mkl-dnn/releases 中对于你的系统的版本,解压后把lib文件夹的 .dylib 文件拷贝到 libtorch/lib/ 里。