Ubuntu20.04上C++部署Pytorch(LibTorch)，OpenCV和ffmpeg避坑安装

安装libtorch：

主要目的是想用C++在服务器上部署深度学习模型，libtorch安装较简单，按步骤既可以
主要注意版本问题，最开始我安装了cu118的最新版本，各种奇怪的报错，后来果断卸载安装cu117,2.0.1对应pytorch版本，就直接安装成功了

参考链接: C++部署pytorch模型
参考链接: Ubuntu 20.04下c++ libtorch gpu配置与运行
部分命令如下：

mkdir build
cd build
cmake -DCMAKE_PREFIX_PATH="/data/l50037097/libtorch" -DCMAKE_CUDA_ARCHITECTURES="80"  ..
make

安装OpenCV:

注意这里不能安装OpenCV 4.X版本 容易和libtoch冲突
安装Opencv 3.4版本即可
参考链接: Ubuntu20安装OpenCV3（图解亲测）

OpenCV和ffmpeg版本也冲突:

解决办法: 直接禁用ffmpeg 后续也不涉及视频处理
链接: opencv+ffmpeg环境（ubuntu）搭建全面详解
链接: link

最后解决

记录最终的CMakeLists.txt文件

cmake_minimum_required(VERSION 3.0)

add_definitions(-D_GLIBCXX_USE_CXX11_ABI=0)

project(cmake_main)

link_directories(/usr/local/cuda-11.7/lib64)  # 替换成您的 CUDA 安装路径

set(CMAKE_CUDA_COMPILER /usr/local/cuda-11.7/bin/nvcc)

set(Torch_DIR /data/l50037097/libtorch/share/cmake/Torch)

find_package(OpenCV REQUIRED)

find_package(Torch REQUIRED)

set(CMAKE_CXX_STANDARD 17)

set(CMAKE_CXX_STANDARD_REQUIRED TRUE)


# include_directories(${OpenCV_INCLUDE_DIRS})

## set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${TORCH_CXX_FLAGS}")

enable_language(CUDA)  # 启用CUDA语言支持

add_executable(cmake_main main_cpp.cpp)

target_link_libraries(cmake_main "${TORCH_LIBRARIES}" ${OpenCV_LIBS})

set_property(TARGET cmake_main PROPERTY CXX_STANDARD 17)

---

结果记录：

torch_test.cpp

#include <torch/script.h>

#include <iostream>
#include <memory>

int main(int argc, const char* argv[]){
    if(argc != 2){
        std::cerr << "usage: main <path-to-exported-script-module>\n";
        return -1;
    }

    torch::Device device(torch::kCUDA);
    // Deserialize the ScriptModule from a file using torch::jit::load()
    torch::jit::script::Module module = torch::jit::load(argv[1]);
    module.to(device);

    // Create a vector of inputs
    std::vector<torch::jit::IValue> inputs;
    inputs.push_back(torch::ones({1, 3, 224, 224}).to(device));

    // Exectute the model
    at::Tensor output = module.forward(inputs).toTensor();

    std::cout << output.slice(/*dims=*/1, /*start=*/0, /*end=*/5) << '\n';

    std::cout << "ok\n";
}

my_test.cpp

#include <torch/script.h>
#include <iostream>
#include <opencv2/opencv.hpp>
#include <torch/torch.h>

// 有人说调用的顺序有关系，我这好像没啥用~~

int main()
{
    torch::DeviceType device_type;
    if (torch::cuda::is_available()) {
        std::cout << "CUDA available! Predicting on GPU." << std::endl;
        device_type = torch::kCUDA;
    }
    else {
        std::cout << "Predicting on CPU." << std::endl;
        device_type = torch::kCPU;
    }
    torch::Device device(device_type);

    //Init model
    std::string model_pb = "tests.pth";
    auto module = torch::jit::load(model_pb);
    module.to(at::kCUDA);

    auto image = cv::imread("dog.jpg", cv::ImreadModes::IMREAD_COLOR);

    cv::Mat image_transfomed;
    cv::resize(image, image_transfomed, cv::Size(32, 32));

    // convert to tensort
    torch::Tensor tensor_image = torch::from_blob(image_transfomed.data,
        { image_transfomed.rows, image_transfomed.cols,3 }, torch::kByte);
    tensor_image = tensor_image.permute({ 2,0,1 });
    tensor_image = tensor_image.toType(torch::kFloat);
    tensor_image = tensor_image.div(255);
    tensor_image = tensor_image.unsqueeze(0);
    tensor_image = tensor_image.to(at::kCUDA);
    torch::Tensor output = module.forward({ tensor_image }).toTensor();
    auto max_result = output.max(1, true);
    auto max_index = std::get<1>(max_result).item<float>();
    std::cout << output << std::endl;
    //return max_index;
    return 0;

}

编译成功！！！！！