ubuntu18.04安装多版本cuda(10.0, 10.1)

当前版本信息
系统: ubuntu18.04
显卡型号: Tesla V100
显卡驱动: NVIDIA-SMI 410.48
cuda版本: 10.0, V10.0.130

打算新安装个cuda10.1版本, 由于显卡驱动版本太低, cuda10.1要求的最低显卡驱动版本是418.39, 所以需要重装显卡驱动,新装cuda10.1和相对应cudnn, 显卡驱动版本对应关系见链接https://docs.nvidia.com/cuda/cuda-toolkit-release-notes/index.html

1. 官网下载显卡驱动和cuda,cudnn
   显卡驱动版本: 418.197.02
   cuda版本: 10.1
   cudnn版本: cuDNN v8.0.5

2. 卸载显卡驱动
   (1) 关掉图形界面

   sudo service lightdm stop
   

   (2) 卸载原有驱动, 一路默认

   sudo /usr/bin/nvidia-uninstall
   

   (3) 重启

   sudo reboot
   

3. 安装新的显卡驱动
   (1) 关掉图形界面

   sudo service lightdm stop
   

   (2) 赋予权限

   chmod +x NVIDIA-Linux-x86_64-418.197.02.run
   

   (3) 安装新的驱动驱动

    ```bash
    sudo ./NVIDIA-Linux-x86_64-418.197.02.run -no-x-check -no-nouveau-check -no-opengl-files
    ```
     *-no-x-check安装驱动时关闭x服务;
    -no-nouveau-check 安装驱动时禁用Nouveau
    -no-opengl-files 安装时只装驱动文件，不安装Opengl*
   

   Would you like to register the kernel module sources with DKMS? This will allow DKMS to automatically build a new module, if you install a different kernel later.
   选择no
   
   WARNING: Unable to find a suitable destination to install 32-bit compatibility libraries. Your system may not be set up for 32-bit compatibility. 32-bit compatibility files will not be
   installed; if you wish to install them, re-run the installation and set a valid directory with the --compat32-libdir option.
   选择 OK
      
   Installation of the kernel module for the NVIDIA Accelerated Graphics Driver for Linux-x86_64 (version 418.197.02) is now complete.
   选择 OK
   

   (4) 启动显示服务

   sudo service lightdm restart
   

   (5) 查看是否安装成功：

   nvidia-smi
   

   

4. 安装cuda10.1
   (1) 安装cuda10.1

   sudo sh cuda_10.1.243_418.87.00_linux.run
   

   安装过程中的选择
   Do you accept the above  EULA? 
   选择accept
   

   

   不修改原有的软链接
   A symlink already exists at /usr/local/cuda. Update to this installation?
   选择No
   

   安装完成后
   此时/burr/local/下

   administrator@dell:/usr/local$ nvcc -V
   nvcc: NVIDIA (R) Cuda compiler driver
   Copyright (c) 2005-2018 NVIDIA Corporation
   Built on Sat_Aug_25_21:08:01_CDT_2018
   Cuda compilation tools, release 10.0, V10.0.130
   administrator@dell:/usr/local$ ls -l
   total 68
   lrwxrwxrwx  1 root root    20 May 14  2020 cuda -> /usr/local/cuda-10.0
   drwxr-xr-x 19 root root  4096 May 14  2020 cuda-10.0
   drwxr-xr-x 14 root root  4096 Jun 24 14:21 cuda-10.1
   

   (2) 在~/.bashrc中添加/修改环境变量：

   sudo vim ~/.bashrc
   

   #将原来的cuda环境变量
   #export PATH=$PATH:/usr/local/cuda-10.0/bin:/home/administrator/anaconda3/bin:/home/administrator/pycharm-2020.1.1/bin
   #export LD_LIBRARY_PATH=/usr/local/cuda-10.0/lib64:$LD_LIBRARY_PATH
   #将原来的cuda-10.0改为cuda
   export PATH=$PATH:/usr/local/cuda/bin:/home/administrator/anaconda3/bin:/home/administrator/pycharm-2020.1.1/bin
   export LD_LIBRARY_PATH=/usr/local/cuda/lib64:$LD_LIBRARY_PATH
   

   在终端运行, 使环境变量生效

   source ~/.bashrc
   

5. 切换cuda版本
   (1) 查看当前cuda版本,为10.0

   administrator@dell:~$ nvcc -V
   nvcc: NVIDIA (R) Cuda compiler driver
   Copyright (c) 2005-2018 NVIDIA Corporation
   Built on Sat_Aug_25_21:08:01_CDT_2018
   Cuda compilation tools, release 10.0, V10.0.130
   

   (2) 切换/usr/local/cuda软链接的指向

   sudo rm -rf /usr/local/cuda  #删除之前创建的软链接
   sudo ln -s /usr/local/cuda-10.1  /usr/local/cuda  #创建新的软链接
   

   (3) 查看cuda版本,发现cuda版本未切换

   dministrator@dell:/usr/local$ nvcc -V
   nvcc: NVIDIA (R) Cuda compiler driver
   Copyright (c) 2005-2018 NVIDIA Corporation
   Built on Sat_Aug_25_21:08:01_CDT_2018
   Cuda compilation tools, release 10.0, V10.0.130
   

   (4) 查看/usr/local/cuda/version.txt中的版本信息

   administrator@dell:/usr/local$ cat /usr/local/cuda/version.txt
   CUDA Version 10.1.243	
   

   (5) 解决方案

   sudo vim /etc/profile
   

   末尾加上

   export PATH=/usr/local/cuda/bin:$PATH
   export LD_LIBRARY_PATH=/usr/local/cuda/lib64$LD_LIBRARY_PATH
   export CUDA_HOME=/usr/local/cuda	
   

   在终端运行, 使环境变量生效

   source /etc/profile
   

   查看cuda版本,此时切换成功

   administrator@dell:/usr/local$ nvcc -V
   nvcc: NVIDIA (R) Cuda compiler driver
   Copyright (c) 2005-2019 NVIDIA Corporation
   Built on Sun_Jul_28_19:07:16_PDT_2019
   Cuda compilation tools, release 10.1, V10.1.243
   

6. 安装cuDNN
   (1) 查看当前cudnn

   administrator@dell:~$ cat /usr/local/cuda/include/cudnn.h | grep CUDNN_MAJOR -A 2
   #define CUDNN_MAJOR 7
   #define CUDNN_MINOR 6
   #define CUDNN_PATCHLEVEL 5
   --
   #define CUDNN_VERSION (CUDNN_MAJOR * 1000 + CUDNN_MINOR * 100 + CUDNN_PATCHLEVEL)
   
   #include "driver_types.h"
   

   (2) 解压cudnn压缩包

   tar -zxf cudnn-10.1-linux-x64-v8.0.5.39.tgz 
   

   (3) 将解压出来的cuda文件夹中的指定文件文件复制到/usr/local/cuda-10.1对应的文件夹下

   sudo cp cuda/include/cudnn*.h /usr/local/cuda-10.1/include
   sudo cp cuda/lib64/libcudnn* /usr/local/cuda-10.1/lib64 
   sudo chmod a+r /usr/local/cuda-10.1/include/cudnn*.h /usr/local/cuda-10.1/lib64/libcudnn*
   

   (4) 验证是否安装成功

   administrator@dell:~$ cat /usr/local/cuda/include/cudnn_version.h | grep CUDNN_MAJOR -A 2
   #define CUDNN_MAJOR 8
   #define CUDNN_MINOR 0
   #define CUDNN_PATCHLEVEL 5
   --
   #define CUDNN_VERSION (CUDNN_MAJOR * 1000 + CUDNN_MINOR * 100 + CUDNN_PATCHLEVEL)
   
   #endif /* CUDNN_VERSION_H */
   

参考文章
[1] 安装多个版本的cuda和cudnn
[2] 非root用户在linux下安装多个版本的CUDA和cuDNN（cuda 8、cuda 10.1 等）
[3] Ubuntu 多版本Cuda(8.0,9.0)以及CuDnn安装