ubuntu16.04-和ubuntu 18.04 安装NVIDIA驱动 GPU

柔情只为你懂 2023-06-11 09:30 10阅读 0赞

ubuntu 16.04 和 18.0——安装CUDA

根据官方文档：[https://docs.nvidia.com/cuda/cuda-installation-guide-linux/index.html][https_docs.nvidia.com_cuda_cuda-installation-guide-linux_index.html]

观察当前系统的显卡驱动 NV132

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70][]

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 1][]

需要停止 Nouveau驱动

[https://docs.nvidia.com/cuda/cuda-installation-guide-linux/index.html][https_docs.nvidia.com_cuda_cuda-installation-guide-linux_index.html]

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 2][]

sudo gedit /etc/modprobe.d/blacklist-nouveau.conf

在文件中加入以下命令：

blacklist nouveau

options nouveau modeset=0

sudo update-initramfs -u

重启电脑：

reboot

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 3][]

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 4][]

\-----------

进入tty模式:

ubuntu 16.04 是 按 Ctrl + Alt + F1键 进入tty模式

ubuntu 18.04是按 Ctrl + Alt + F2 （tty2） Ctrl + Alt +F3 (tty3) 直到 Ctrl +Alt+ F6(tty6)都可以

\-----------------------------------------------------------

ubuntu 18.04 默认没有安装 build-essential

ubuntu 16.04 默认是安装 build-essential的

\-------------------------------------------------------------

卸载已安装的CUDA程序（安装是10.1版本为例）

sudo /usr/local/cuda-10.1/bin/cuda-uninstaller

过程如下：

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 5][]

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 6][]

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 7][]

卸载成功后，会提醒卸载成功

![20191030093423583.png][]

卸载已安装的 nvidia驱动

nvidia驱动的卸载程序安装在 /usr/bin下：

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 8][]

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 9][]

点击“yes”卸载完成，即可。

ubuntu 16.04

安装驱动

安装nvidia驱动，会出现以下错误无法安装，根据提示发现没有关闭图形界面

![20191030100829552.png][]

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 10][]

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 11][]

正确方式如下：

按下 Ctrl+Alt+F1 进入 tty1模式：

关闭界面

sudo /etc/init.d/lightdm stop

或者运行以下脚本

sudo service lightdm stop

然后才可以安装，安装命令如下：

sudo ./NVIDIA-Linux-x86_64-430.50.run

顺利安装过程完后，

sudo /etc/init.d/lightdm start

或者

sudo service lightdm start

按键 Ctrl+Alt+F7，进入图形界面。

ubuntu 18.04安装nvidia驱动

ubuntu 18.04是没有安装 'build-essential'，首先安装gcc，否则在安装nvidia驱动时，会出现错误。

sudo apt-get install build-essentail

刚开始以为和ubuntu 16.4是一样的，结果不用关系图形界面，即不用以下命令

sudo /etc/init.d/gdm3 stop

相当于 16.4系统的

sudo /etc/init.d/lightdm stop

可以直接安装，即：

sudo ./NVI   .run

安装nvidia 是需要 gcc的，否则会错误。

安装结束后，安装路径为 /usr/bin/

ubuntu18.04 系统安装 CUDA-10.1 是要关闭图形界面的，

\---------------------------------------------------------------------------------

如果没有禁用nouveau开源驱动，直接运行以下脚本安装

sudo ./NVIDIA-Linux-x86_64-430.50.run

会报以下错误，提示现在系统正在使用 nouveau 驱动，需要禁止 nouveau 驱动

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 12][]

\---------------------

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 13][]

\-----

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 14][]

\-----

如何停止nouveau呢。如果运行了以上步骤，会在/etc/modprobe.d 文件夹中，自动产生以下文件：

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 15][]

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 16][]

以上为提示。以下运行脚本停止nouveau 驱动：

，停止nouveau 驱动：

sudo gedit /etc/modprobe.d/blacklist.conf

在文件中加入：

blacklist nouveau

options nouveau modeset=0

运行以下脚本：

sudo update-initramfs -u

输出结果：

Generating /boot/initrd.img-5.0.0-32-generic

【1：initramfs : initram file system 内存文件系统，initramfs】

【2：update-initramfs 该命令位于/usr/sbin下，update-initramfs --h 换取帮助命令】

reboot | 安装 nvidia驱动之前，还需要安装gcc，需要编译。否则，也会出现以下错误：

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 17][]

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 18][]

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 19][]

sudo apt-get install build-essential

\-----------

1：禁用nouveau驱动

2：安装build-essential之后

然后就可以安装nvidia驱动了【注意，可以不像ubuntu16下，必须在tty模式和禁用图形界面后，才可以安装】

sudo ./NVIDIA-Linux-x86_64-430.50.run

安装完成后，就可以重启了。

reboot

查看nvidia驱动的信息，

nvidia-smi

\------------

出现了四个进程的原因，有可能在tty2中操作安装了 nvidia驱动，Ctrl+Alt+F

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 20][]

出现以上情况，发现进程很难杀死，

sudo killall -9 /usr/bin/gnome-shell

sudo killall -9 /usr/lib/xorg/Xorg

操作步骤如下：

通过以下命令进入tty模式，

Ctrl+Alt+F2或者F3,

进入root用户或者sudo 操作，运行以下脚本：

sudo /etc/init.d/gdm3 stop

这时候，然后再运行

nvidia-smi

发现杀死了两个进程。

或者进入tty模式，然后利用kill 杀死两个进程，然后执行以下脚本

/etc/init.d/gdm3 stop

然后再利用

然后再利用kill或者killall来逐个杀死GPU的进程。

sudo ./cuda_10.1.105_418.39_linux.run

会提示安装成功。

\-----------------

测试如果不杀死GPu进程来安装 ，会在文件夹/var/log下产生一个cuda-installer.log日志，该日志记录错误信息

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 21][]

查看以上日志，找到错误如下。

[INFO]: ERROR: An NVIDIA kernel module 'nvidia-drm' appears to already be loaded in
    
    [INFO]:        your kernel.  This may be because it is in use (for example, by an X
    
    [INFO]:        server, a CUDA program, or the NVIDIA Persistence Daemon), but this
    
    [INFO]:        may also happen if your kernel was configured without support for
    
    [INFO]:        module unloading.  Please be sure to exit any programs that may be
    
    [INFO]:        using the GPU(s) before attempting to upgrade your driver.  If no
    
    [INFO]:        GPU-based programs are running, you know that your kernel supports
    
    [INFO]:        module unloading, and you still receive this message, then an error
    
    [INFO]:        may have occured that has corrupted an NVIDIA kernel module's usage
    
    [INFO]:        count, for which the simplest remedy is to reboot your computer.
    
    [INFO]: 
    
    [INFO]: 
    
    [INFO]: ERROR: Installation has failed.  Please see the file
    
    [INFO]:        '/var/log/nvidia-installer.log' for details.

\------------------------------------------------------------------------------------------------------------------------------------------------------------------

安装CUDA-10时，经常出现一些问题：

\--------------------------------------------------------------------

ubuntu 16.04时，利用runfile安装 CUDA10.0时，始终安装不成功，利用deb安装成功。

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 22][]

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 23][]

下载完成后，看到一个后缀为.deb的文件

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 24][]

安装完成后，配置环境变量

LD\_LIBRARY\_PATH=/usr/local/cuda/lib64:$LD\_LIBRARY\_PATH

export LD\_LIBRARY\_PATH

PATH=/usr/local/cuda/bin:$PATH

export PATH

————————————————————————————————————————

[https://www.lfd.uci.edu/~gohlke/pythonlibs/][https_www.lfd.uci.edu_gohlke_pythonlibs]

[https_docs.nvidia.com_cuda_cuda-installation-guide-linux_index.html]: https://docs.nvidia.com/cuda/cuda-installation-guide-linux/index.html
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70]: /images/20230531/9addb2bced1946f39372217725fe74dd.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 1]: /images/20230531/fdfd4faaeba24bc793941a65fe4285bf.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 2]: /images/20230531/8763548f0a134a01aa112f2a5952f45a.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 3]: /images/20230531/50ce39f1c5ad4534ae57661dc8de6fc2.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 4]: /images/20230531/54a6d17f371b4db1b67ebec9e664568e.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 5]: /images/20230531/3af220b0c8db46368f6168f4fbc97b7b.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 6]: /images/20230531/2ef7175b5193425c9c7ef4bc3c57d295.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 7]: /images/20230531/a85cc9f59b7b48a2a131f2a058173e0c.png
[20191030093423583.png]: /images/20230531/562edf55d9d745fe881c75af57491688.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 8]: /images/20230531/307f0666246444b0be65ddc52a89310c.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 9]: /images/20230531/4cedb39048474c5dade1d827c1e99aa3.png
[20191030100829552.png]: /images/20230531/c545de5f82b348edb092086f3f99568e.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 10]: /images/20230531/e5786ba1fb57474cb421386c62d4cd4b.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 11]: /images/20230531/afa5dc8834fa4bf381128bb653f32439.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 12]: /images/20230531/43ef5808889c4856be3fe8507e2c22d0.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 13]: /images/20230531/7ff172f3d13b474ca1d7254a682709c2.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 14]: /images/20230531/71a4a131619a4b1fba98d51dbc352269.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 15]: /images/20230531/e6604345269d40f9aa427c57e0077fa1.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 16]: /images/20230531/e0967c52bbbe489fba63723887ee90ff.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 17]: /images/20230531/a39f2b7195bc4158a4fb57e9d9865550.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 18]: /images/20230531/caad197eb6ba404ca12c34f7657ee2e8.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 19]: /images/20230531/93112fd3f6074df89916524618eb094c.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 20]: /images/20230531/63be7d10388e47108860616a59285c46.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 21]: /images/20230531/a17f6c1b88ac4984b8ab698e63c20f51.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 22]: /images/20230531/a7076818739849a58386c8e4f928d890.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 23]: /images/20230531/f47098a8e8c441ac9d8720a2d2e00d50.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly96aGFuZ3BpbmcuYmxvZy5jc2RuLm5ldA_size_16_color_FFFFFF_t_70 24]: /images/20230531/e50a3b95bc904891a8662a0a881fef83.png
[https_www.lfd.uci.edu_gohlke_pythonlibs]: https://www.lfd.uci.edu/~gohlke/pythonlibs/