1. 安装方式

1. 在线安装

# 默认安装位置：/usr/bin/
sudo apt-get install nvidia-container-runtime

2. 离线安装

1. 获取离线安装包。

• 域名配置 ```bash

  通过域名查询IP

  https://ip.tool.chinaz.com/

  https://site.ip138.com/

配置域名解析

vi /etc/hosts

添加以下条目

185.199.109.153 nvidia.github.io

- **安装apt-transport-https**
```bash
wget http://ftp.hk.debian.org/debian/pool/main/a/apt/apt-transport-https_1.0.9.8.4_amd64.deb
dpkg -i apt-transport-https_*_amd64.deb

• 软件源配置

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

• 联网下载安装包

  apt-get download libnvidia-container1
  apt-get download libnvidia-container-tools
  apt-get download nvidia-container-runtime-hook
  apt-get download nvidia-container-runtime
  apt-get download nvidia-docker2

1. 离线安装。

   # 注意：顺序不能颠倒
   dpkg -i libnvidia-container1_1.0.5-1_amd64.deb
   dpkg -i libnvidia-container-tools_1.0.5-1_amd64.deb
   dpkg -i nvidia-container-runtime-hook_1.4.0-1_amd64.deb
   dpkg -i nvidia-container-runtime_2.0.0+docker18.06.2-1_amd64.deb
   dpkg -i nvidia-docker2_2.0.3+docker18.06.2-1_all.deb

   2. 配置

   # 配置默认nvidia运行（安装成功或一般会默认安装上）
   sudo vi /etc/docker/daemon.json

   内容如下：

   {
   "default-runtime": "nvidia"
    "runtimes": {
        "nvidia": {
            "path": "/usr/bin/nvidia-container-runtime",
            "runtimeArgs": []
        }
    }
   }

   重启Docker：

   # 重启docker服务即可生效
   sudo systemctl daemon-reload
   sudo systemctl restart docker

   3. 验证

   sudo docker run --runtime=nvidia --rm hello-world
   sudo docker run --runtime=nvidia --rm nvidia/cuda:9.0-base nvidia-smi

   4. 移除

   sudo apt-get purge libnvidia-container1
   sudo apt-get purge libnvidia-container-tools
   sudo apt-get purge nvidia-container-runtime-hook
   sudo apt-get purge nvidia-container-runtime
   sudo apt-get purge nvidia-docker2

   参考

   博文：docker 和 nvidia-docker 的离线安装和基本使用
   https://fanfuhan.github.io/2019/11/22/docker_based_use/