安装显卡驱动、安装cuda toolkit、安装cudnn

nvidia-smi查看，CUDA可以向下兼容，以我为例，可以安装12.0以下的版本：

cuda toolkit

https://developer.nvidia.com/cuda-toolkit-archive

如果 你前面没有安装vs，直接安装的这个，需要把CUDA里面的
Visual Studio Integration取消勾选，否则会安装不成功

新建文件夹
D:\cuda11.6\Documentation
D:\cuda11.6\Samples
D:\cuda11.6\Development

Installed:
- Nsight Monitor
Not Installed:
- Nsight for Visual Studio 2019
Reason: VS2019 was not found
- Nsight for Visual Studio 2017
Reason: VS2017 was not found
- Integrated Graphics Frame Debugger and Profiler
Reason: see https://developer.nvidia.com/nsight-vstools
- Integrated CUDA Profilers
Reason: see https://developer.nvidia.com/nsight-vstools

打开cmd，输入 nvcc -V
出现了你安装的CUDA的版本信息，说明安装成功了

cudnn

https://developer.nvidia.com/rdp/cudnn-archive

为下载cudnn创建了NVIDIA账号记录：有账号就不管

把这三个文件夹的文件分别拷贝到CUDA安装目录对应的（bin、include、lib）文件夹中即可。CUDA的lib目录有x64 、Win32、cmake三个文件夹，拷到其中的x64这个文件夹中

都装好了

配置环境变量

注意：如果需要管理员权限，则可通过修改属性，提供权限

没成功-没权限

D:\cuda11.6\Development\bin
D:\cuda11.6\Development\include
D:\cuda11.6\Development\lib
D:\cuda11.6\Development\libnvvp

Path后面添加
;%D:\cuda11.6\Development\bin;%D:\cuda11.6\Development\include;%D:\cuda11.6\Development\lib;%D:\cuda11.6\Development\libnvvp

参考：https://blog.csdn.net/ZHUO__zhuo/article/details/130536374
https://blog.csdn.net/chen565884393/article/details/127905428
https://blog.csdn.net/qq_46941656/article/details/119681765