前言

博主由于视觉开发需求, 配置nvidia驱动并映射到docker中运行, 在本文中记录过程及遇到的问题

硬件及软件环境

   Static hostname: debian
         Icon name: computer-desktop
           Chassis: desktop
  Operating System: Debian GNU/Linux 11 (bullseye)
            Kernel: Linux 5.10.0-19-amd64
      Architecture: x86-64
     
  CPU: 12th Gen Intel(R) Core(TM) i7-12700F
  GPU: Nvidia Quadro M2000

驱动

下载驱动

根据自己的显卡型号去官网搜索对应的驱动程序. 本机选择470.161…03版本驱动.
NVIDIA 驱动程序下载 官方高级驱动搜索

cuda对应驱动版本要求对照表:
NVIDIA CUDA Toolkit Release Notes

注意!

1. 直接使用apt-get install nvidia-driver时不可运行(can not communicate with nvidia driver 类似报错)
2. 下载最新驱动525时不可运行(can not communicate with nvidia driver 类似报错)
3. 安装时需屏蔽x server及nouveau¹

安装

禁用xserver

sudo service gdm3 stop

输入该行指令后会进入命令行状态, 此时只有一个光标, 通过按Ctrl + Alt + F1和Ctrl + Alt + F2即可跳出输入用户名密码的指令行.

禁用nouveau

sudo bash -c "echo blacklist nouveau > /etc/modprobe.d/blacklist-nvidia-nouveau.conf"
sudo bash -c "echo options nouveau modeset=0 >> /etc/modprobe.d/blacklist-nvidia-nouveau.conf"

禁用后重启计算机

sudo reboot

安装依赖

后续编译使用的依赖软件²

sudo apt-get install gcc g++ cmake pkg-config
sudo apt-get install linux-headers-$(uname -r|sed 's/[^-]*-[^-]*-//')

设置可执行并运行

chmod +x ~/Downloads/NVIDIA-Linux-x86_64-470.161.03.run

# 需要以管理员权限运行
sudo ~/Downloads/NVIDIA-Linux-x86_64-470.161.03.run 

安装完成后, 重启计算机并删除禁用nouveau时创建的blacklist文件

检查安装结果

nvidia-smi

# 输出
Thu Mar  9 14:22:29 2023       
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 470.161.03   Driver Version: 470.161.03   CUDA Version: 11.4     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  Quadro M2000        Off  | 00000000:01:00.0  On |                  N/A |
| 63%   59C    P0    38W /  75W |    769MiB /  4041MiB |     30%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
|    0   N/A  N/A      1835      G   /usr/lib/xorg/Xorg                282MiB |
|    0   N/A  N/A      1982      G   /usr/bin/gnome-shell              110MiB |
|    0   N/A  N/A     30799      G   gnome-control-center               39MiB |
+-----------------------------------------------------------------------------+

Docker配置

Docker安装

安装可以参考此文如何建立并使用docker

nvidia-container-runtime安装³

命令

nano nvidia-container-runtime-script.sh

脚本内容

curl -s -L https://nvidia.github.io/nvidia-container-runtime/gpgkey | \
  sudo apt-key add -
distribution=$(. /etc/os-release;echo $ID$VERSION_ID)
curl -s -L https://nvidia.github.io/nvidia-container-runtime/$distribution/nvidia-container-runtime.list | \
  sudo tee /etc/apt/sources.list.d/nvidia-container-runtime.list
sudo apt-get update

执行脚本

nvidia-container-runtime-script.sh

安装 nvidia-container-runtime

sudo apt-get install nvidia-container-runtime
sudo systemctl restart docker # 重启docker

检测

which nvidia-container-runtime-hook 
/usr/bin/nvidia-container-runtime-hook

Docker gpu 验证

docker pull nvidia/cuda:11.3.1-base-ubuntu20.04
docker run --gpus all --rm -it nvidia/cuda:11.3.1-base-ubuntu20.04 bash

nvidia-smi
#输出如下, 说明运行成功:
root@8a57ae3075d7:/# nvidia-smi
Thu Mar  9 06:42:20 2023       
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 470.161.03   Driver Version: 470.161.03   CUDA Version: 11.4     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  Quadro M2000        Off  | 00000000:01:00.0  On |                  N/A |
| 62%   53C    P0    28W /  75W |    761MiB /  4041MiB |     34%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
+-----------------------------------------------------------------------------+

卸载指令

卸载安装的驱动可以使用⁴:

sudo apt-get --purge remove "*nvidia*"
sudo /usr/bin/nvidia-uninstall

总结

本文用来记录Debian11在安装nvidia驱动和docker运行时遇到的一些问题, 由于是事后补写可能中间有些异常处理略有缺漏, 各位同学有问题可以留言交流.

异常处理参考链接

显卡驱动报错：NVIDIA-SMI has failed because it couldn’t communicate with the NVIDIA driver.
固件错误Possible missing firmware解决办法

---

1. Debian安装英伟达(NVIDIA)驱动一站式避坑教学(Ubuntu通用) ↩︎

2. Debian 10.2命令安装Nvidia显卡驱动成功，问题回顾 ↩︎

3. Docker GPU 调用 ↩︎

4. Ubuntu 卸载 Nvidia 驱动和安装最新驱动 ↩︎