报错解决：RuntimeError:The detected CUDA version mismatches the version that was used to compile PyTorch.
报错解决：RuntimeError:The detected CUDA version mismatches the version that was used to compile PyTorch.
- 报错
- 解决方法：
-
- cuda安装
- cuda多版本切换
-
- 方法一：通过修改软链接的方式
- 方法二：修改bashrc中cuda的路径
- 附录：可能遇到的报错
-
- 报错一：GCC版本不兼容
- 报错二：安装路径报错
报错
博主在编译安装软件时，遇到报错如下：
File "/home/XXX/miniconda3/envs/lin/lib/python3.8/site-packages/torch/utils/cpp_extension.py", line 404, in build_extensions
self._check_cuda_version()
File "/home/XXX/miniconda3/envs/lin/lib/python3.8/site-packages/torch/utils/cpp_extension.py", line 781, in _check_cuda_version
raise RuntimeError(CUDA_MISMATCH_MESSAGE.format(cuda_str_version, torch.version.cuda))
RuntimeError:
The detected CUDA version (11.3) mismatches the version that was used to compile PyTorch (10.2). Please make sure to use the same CUDA versions.
报错原因：CUDA版本和Pytorch版本不匹配。
解决方法：
cuda安装
前往官网，下载低版本的cuda并安装，例如我选择安装cuda10.2，网页如下图所示：
运行如下指令进行安装：
wget https://developer.download.nvidia.com/compute/cuda/10.2/Prod/local_installers/cuda_10.2.89_440.33.01_linux.run
sudo sh cuda_10.2.89_440.33.01_linux.run
输入accept：
取消安装cuda驱动（因为在这之前已经安装好更高版本的显卡驱动就不需要再重复安装）
Do you want to install a symbolic link at /usr/local/cuda? # 是否将安装目录通过软连接的方式 link 到 /usr/local/cuda
选择yes
安装完成后，终端会有如下输出：
===========
= Summary =
===========
Driver: Not Selected
Toolkit: Installed in /usr/local/cuda-10.2/
Samples: Installed in /root/, but missing recommended libraries
Please make sure that
- PATH includes /usr/local/cuda-10.2/bin
- LD_LIBRARY_PATH includes /usr/local/cuda-10.2/lib64, or, add /usr/local/cuda-10.2/lib64 to /etc/ld.so.conf and run ldconfig as root
To uninstall the CUDA Toolkit, run cuda-uninstaller in /usr/local/cuda-10.2/bin
Please see CUDA_Installation_Guide_Linux.pdf in /usr/local/cuda-10.2/doc/pdf for detailed information on setting up CUDA.
***WARNING: Incomplete installation! This installation did not install the CUDA Driver. A driver of version at least 440.00 is required for CUDA 10.2 functionality to work.
To install the driver using this installer, run the following command, replacing <CudaInstaller> with the name of this run file:
sudo <CudaInstaller>.run --silent --driver
Logfile is /var/log/cuda-installer.log
检查cuda版本：
nvcc --version
结果如上图所示，说明安装成功。
cuda多版本切换
如果之后需要进行其他cuda版本的切换可以按照如下两个方法：
方法一：通过修改软链接的方式
- 将~/.bashrc 下与cuda相关的路径都改为/usr/local/cuda/，而不使用具体某一cuda版本，例如cuda-10.2、cuda-11.3等。
- 修改软连接
# 删除之前创建的软链接
rm -rf /usr/local/cuda
# 建立新的软连接
sudo ln -s /usr/local/cuda-11.3/ /usr/local/cuda/
# 查看当前cuda版本
nvcc --version
方法二：修改bashrc中cuda的路径
打开bashrc文件，在最后添加如下：（根据需要使用cuda版本修改版本号）：
export PATH="/usr/local/cuda-10.2/bin:$PATH"
export LD_LIBRARY_PATH="/usr/lcoal/cuda-10.2/lib64:$LD_LIBRARY_PATH"
之后source更新一下source ~/.bashrc。
附录：可能遇到的报错
报错一：GCC版本不兼容
Failed to verify gcc version. See log at /var/log/cuda-installer.log for details.
使用以下指令进行安装：
sudo sh cuda_10.2.89_440.33.01_linux.run --override
报错二：安装路径报错
Installation failed. See log at /var/log/cuda-installer.log for details.
查看log的报错原因
cat /var/log/cuda-installer.log | grep [ERROR]
发现报错原因是：
[ERROR]: boost::filesystem::remove: Directory not empty: "/var/log/nvidia/.uninstallManifests/"
使用以下命令进行安装：
sudo sh cuda_10.2.89_440.33.01_linux.run --librarypath=/usr/local/cuda-10.2