cuda编译报错

1.报错：

CUDA_ARCHITECTURES is empty for target "cmTC_d5b93".

解决：

cmake .. -DUSE_CUDA=ON -DCMAKE_CUDA_ARCHITECTURES=86 -DCMAKE_BUILD_TYPE=Release

2.
报错：ptxas fatal : Value 'sm_30' is not defined for option 'gpu-name'

查看: /usr/bin/nvcc 版本

3.NvInferRuntime.h: 没有那个文件或目录
添加tensorrt头文件目录
list(APPEND IncludeList "/usr/local/TensorRT-8.4.3.1/include")
或者
include_directories(
    ${PROJECT_SOURCE_DIR}/include
    /usr/local/TensorRT-8.4.3.1/include
)

4./usr/bin/ld: 找不到 -lnvinfer
include_directories(/usr/local/TensorRT-8.4.3.1/include) 
link_directories(/usr/local/TensorRT-8.4.3.1/lib)

5.
  Could not find a package configuration file provided by "Torch" with any of
  the following names:

    TorchConfig.cmake
    torch-config.cmake
解决：
下载libtorch  
libtorch-cxx11-abi-shared-with-deps-1.13.1+cu116.zip
并设置cmakelist.txt
set(CMAKE_PREFIX_PATH ${PROJECT_SOURCE_DIR}/thirdparty/libtorch)