1 离线安装Nvidia Driver

参考网址：
Deepin 20.9在GTX 1060显卡上安装Nvidia 550.100驱动

2 离线安装Docker Engine

1. 卸载旧版本包

# 强制删除docker的缓存
docker system prune -a --force

# 卸载docker相关安装包
for pkg in docker.io docker-doc docker-compose docker-compose-v2 podman-docker containerd runc; do sudo apt-get remove $pkg; done

sudo apt-get purge docker-ce docker-ce-cli containerd.io docker-buildx-plugin docker-compose-plugin docker-ce-rootless-extras

sudo rm -rf /var/lib/docker
sudo rm -rf /var/lib/containerd

3. 下载安装软件
   下载地址: https://download.docker.com/linux/debian/dists/bullseye/pool/stable/amd64/

wget https://download.docker.com/linux/debian/dists/bullseye/pool/stable/amd64/containerd.io_1.7.19-1_amd64.deb


wget https://download.docker.com/linux/debian/dists/bullseye/pool/stable/amd64/docker-ce-cli_27.0.3-1~debian.11~bullseye_amd64.deb

wget https://download.docker.com/linux/debian/dists/bullseye/pool/stable/amd64/docker-ce_27.0.3-1~debian.11~bullseye_amd64.deb

wget https://download.docker.com/linux/debian/dists/bullseye/pool/stable/amd64/docker-buildx-plugin_0.15.1-1~debian.11~bullseye_amd64.deb

wget https://download.docker.com/linux/debian/dists/bullseye/pool/stable/amd64/docker-compose-plugin_2.28.1-1~debian.11~bullseye_amd64.deb

4. 安装相关软件

sudo dpkg -i ./containerd.io_1.7.19-1_amd64.deb \
./docker-ce-cli_27.0.3-1~debian.11~bullseye_amd64.deb \
./docker-ce_27.0.3-1~debian.11~bullseye_amd64.deb \
./docker-buildx-plugin_0.15.1-1~debian.11~bullseye_amd64.deb \
./docker-compose-plugin_2.28.1-1~debian.11~bullseye_amd64.deb

5. 添加镜像源

# 添加镜像源
sudo vim /etc/docker/daemon.json

{
    "registry-mirrors": [
        "https://hub.uuuadc.top",
        "https://hub-mirror.c.163.com",
        "https://mirror.baidubce.com",
        "https://docker.anyhub.us.kg",
        "https://dockerhub.jobcher.com",
        "https://dockerhub.icu",
        "https://docker.ckyl.me",
        "https://docker.awsl9527.cn",
        "https://docker.mirrors.ustc.edu.cn",
        "https://docker.nju.edu.cn",
        "https://docker.m.daocloud.io",
        "https://dockerproxy.com",
        "https://registry.docker-cn.com",
        "https://hub-mirror.c.163.com"
    ]
}

# 使配置生效
sudo systemctl daemon-reload

# 重启Docker
sudo systemctl restart docker

5. 添加root权限

sudo groupadd docker          # 添加docker用户组
sudo usermod -aG docker $USER # 将当前用户添加到docker用户组
newgrp docker                 # 更新用户组
docker info                   # 测试docker命令是否可以使用sudo正常使用
reboot                        # 重启系统

6. 验证安装成功

sudo service docker start

sudo docker run --rm hello-world

参考网址：

1. Install Docker Engine on Ubuntu
2. Manage Docker as a non-root user

7 离线安装Nvidia Container Toolkit

1. 下载安装软件
   下载地址: https://github.com/NVIDIA/libnvidia-container/tree/gh-pages/stable/debian11/

# libnvidia-container及其工具包
libnvidia-container1_1.13.5-1_amd64.deb
libnvidia-container-tools_1.13.5-1_amd64.deb

# nvidia-container-toolkit及其基础文件
nvidia-container-toolkit-base_1.13.5-1_amd64.deb
nvidia-container-toolkit_1.13.5-1_amd64.deb

# nvidia-container-runtime
nvidia-container-runtime_3.13.0-1_all.deb

# nvidia-docker2
nvidia-docker2_2.13.0-1_all.deb

# libnvidia-container开发调试包（一般不需要安装）
libnvidia-container-dev_1.13.5-1_amd64.deb
libnvidia-container1-dbg_1.13.5-1_amd64.deb

2. 安装相关软件

# 安装相关软件
sudo dpkg -i ./libnvidia-container1_1.13.5-1_amd64.deb \
./libnvidia-container-tools_1.13.5-1_amd64.deb \
./nvidia-container-toolkit-base_1.13.5-1_amd64.deb \
./nvidia-container-toolkit_1.13.5-1_amd64.deb \
./nvidia-container-runtime_3.13.0-1_all.deb \
./nvidia-docker2_2.13.0-1_all.deb

# 修复依赖关系
sudo apt-get install -f

3. 添加镜像源

# 添加镜像源
sudo vim /etc/docker/daemon.json

{
    "runtimes": {
        "nvidia": {
            "path": "nvidia-container-runtime",
            "runtimeArgs": []
        }
    },
    "registry-mirrors": [
        "https://hub.uuuadc.top",
        "https://hub-mirror.c.163.com",
        "https://mirror.baidubce.com",
        "https://docker.anyhub.us.kg",
        "https://dockerhub.jobcher.com",
        "https://dockerhub.icu",
        "https://docker.ckyl.me",
        "https://docker.awsl9527.cn",
        "https://docker.mirrors.ustc.edu.cn",
        "https://docker.nju.edu.cn",
        "https://docker.m.daocloud.io",
        "https://dockerproxy.com",
        "https://registry.docker-cn.com",
        "https://hub-mirror.c.163.com"
    ]
}

# 使配置生效
sudo systemctl daemon-reload

# 重启Docker
sudo systemctl restart docker

3. 验证安装成功

sudo systemctl restart docker

docker run --gpus all --rm nvcr.io/nvidia/cuda nvidia-smi

参考网址：

1. 配置 Docker 使用 GPU
2. Ubuntu20.04 arm64/amd64 离线安装nvidia-docker2
3. docker 使用gpu
4. libnvidia-container
5. NVIDIA Container Toolkit repository
6. Docker 配置国内镜像源加速