以下环境均为:Ubuntu22.04

查看主机显卡

lspci | grep NVIDIA

安装Nvidia驱动

主机上需要安装Nvidia驱动,可进入Nvidia官网 (opens new window)下载对应系统驱动。

ubuntu 官方安装文档 (opens new window)

version=570.133.20
distro=ubuntu2004
arch_ext=amd64.deb

apt install linux-headers-$(uname -r)
wget https://developer.download.nvidia.com/compute/nvidia-driver/$version/local_installers/nvidia-driver-local-repo-$distro-$version_$arch_ext.deb
dpkg -i nvidia-driver-local-repo-$distro-$version_$arch_ext.deb

# 更新源
apt update
cp /var/nvidia-driver-local-repo-$distro-$version/nvidia-driver-*-keyring.gpg /usr/share/keyrings/

# 查看驱动
apt search nvidia-driver
apt install nvidia-driver-570

或者

## 查看当前系统推荐的 NVIDIA 驱动版本。
sudo ubuntu-drivers devices
## 搭配 autoinstall 自动安装推荐驱动
sudo ubuntu-drivers autoinstall

运行命令查看驱动是否正常安装

nvidia-smi

示例

root@ecs001:~# nvidia-smi
Thu May 22 14:43:38 2025
+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 570.133.20             Driver Version: 570.133.20     CUDA Version: 12.8     |
|-----------------------------------------+------------------------+----------------------+
| 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  Tesla T4                       Off |   00000000:00:07.0 Off |                    0 |
| N/A   34C    P8              9W /   70W |       8MiB /  15360MiB |      0%      Default |
|                                         |                        |                  N/A |
+-----------------------------------------+------------------------+----------------------+

+-----------------------------------------------------------------------------------------+
| Processes:                                                                              |
|  GPU   GI   CI              PID   Type   Process name                        GPU Memory |
|        ID   ID                                                               Usage      |
|=========================================================================================|
|    0   N/A  N/A            1140      G   /usr/lib/xorg/Xorg                        4MiB |
+-----------------------------------------------------------------------------------------+

cuda-toolkit-12-9

提示

深度学习CUDA环境，不需要可以不安装

参考官网 (opens new window)

wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64/cuda-ubuntu2004.pin
sudo mv cuda-ubuntu2004.pin /etc/apt/preferences.d/cuda-repository-pin-600
wget https://developer.download.nvidia.com/compute/cuda/12.9.0/local_installers/cuda-repo-ubuntu2004-12-9-local_12.9.0-575.51.03-1_amd64.deb
sudo dpkg -i cuda-repo-ubuntu2004-12-9-local_12.9.0-575.51.03-1_amd64.deb
sudo cp /var/cuda-repo-ubuntu2004-12-9-local/cuda-*-keyring.gpg /usr/share/keyrings/
# 更新源
sudo apt-get update
sudo apt-get -y install cuda-toolkit-12-9

安装 nvidia-container-toolkit

参考官网 (opens new window)

curl -fsSL https://nvidia.github.io/libnvidia-container/gpgkey | sudo gpg --dearmor -o /usr/share/keyrings/nvidia-container-toolkit-keyring.gpg \
  && curl -s -L https://nvidia.github.io/libnvidia-container/stable/deb/nvidia-container-toolkit.list | \
    sed 's#deb https://#deb [signed-by=/usr/share/keyrings/nvidia-container-toolkit-keyring.gpg] https://#g' | \
    sudo tee /etc/apt/sources.list.d/nvidia-container-toolkit.list

sudo apt-get update

sudo apt-get install -y nvidia-container-toolkit

配置 docker 使用nvidia runtime

sudo nvidia-ctk runtime configure --runtime=docker

sudo systemctl restart docker

查看配置文件

cat /etc/docker/daemon.json

应该出现如下结果

{
    "default-runtime": "nvidia",
    "runtimes": {
        "nvidia": {
            "args": [],
            "path": "nvidia-container-runtime"
        }
    }
}

使用

docker使用

• --gpus all：表示将所有 GPU 都分配给该容器
• --gpus "device=$id,"分配指定gpu

测试

docker run --rm --gpus all  nvidia/cuda:12.8.0-runtime-ubuntu22.04 nvidia-smi

Docker Swarm

services:
  cuda12:
    image: nvidia/cuda:12.8.0-runtime-ubuntu22.04
    container_name: cuda12.8
    restart: "no"
    environment:
      NVIDIA_DRIVER_CAPABILITIES: compute,video,utility
    command: ["nvidia-smi"]
    deploy:
      resources:
        reservations:
          devices:
            - capabilities: [gpu] # 旧 Swarm 用法，

docker-conpose v2使用

services:
  cuda12:
    image: nvidia/cuda:12.8.0-runtime-ubuntu22.04
    container_name: cuda12.8
    restart: "no"
    environment:
      NVIDIA_DRIVER_CAPABILITIES: compute,video,utility # compute:计算  video:视频编解码 utility:工具 all:所有模块
    command: ["nvidia-smi"]
    # runtime: nvidia
    # 新版 Docker Compose 也可以用 device_requests 和 runtime: nvidia 取其一
    device_requests:
      - driver: nvidia
        count: all
        capabilities: ["gpu"]

• runtime: nvidia 告诉 Docker运行时用NVIDIA GPU容器替代默认的runc的runtime。
• device_requests 运行时请求 要暴露哪些设备给容器。

配置默认

提示

docker-conpose配置可省略runtime: nvidia

vi /etc/docker/daemon.json 

添加"default-runtime": "nvidia"，示例如下

{
    "default-runtime": "nvidia",
    "runtimes": {
        "nvidia": {
            "args": [],
            "path": "nvidia-container-runtime"
        }
    }
}