Error response from daemon: could not select device driver ““ with capabilities: [[gpu]]

1. 报错信息

修改Ubuntu驱动后，启动docker容器时报错：

Error response from daemon: could not select device driver "" with capabilities: [[gpu]]
Error: failed to start containers: cu111

2. 确认 NVIDIA 驱动安装

确保 NVIDIA 驱动已正确安装并正常运行。运行以下命令检查：

nvidia-smi

如果能正常显示驱动信息，则驱动已安装。

3. 安装 NVIDIA Docker 运行时

确保安装了 NVIDIA Docker 运行时。使用以下命令安装：

# 添加 NVIDIA Docker 的 GPG 密钥
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

# 更新包信息并安装 NVIDIA Docker
sudo apt update
sudo apt install -y nvidia-docker2

上述命令可能会出现两个报错：

报错1：

curl -s -L https://nvidia.github.io/nvidia-docker/gpgkey | sudo apt-key add -

执行上述命令时，获取失败：

gpg: no valid OpenPGP data found.

解决方法

手动本地下载文件：复制链接到浏览器打开，就会自动下载：

https://nvidia.github.io/nvidia-docker/gpgkey
上传到服务器，执行下面命令，安装，记得替换路径。

 sudo apt-key add "/home/gy77/gpgkey"

报错2：

sudo apt install -y nvidia-docker2

执行上述命令时，nvidia-docker2时报错：

Err:1 https://nvidia.github.io/libnvidia-container/stable/deb/amd64  nvidia-container-toolkit-base 1.15.0-1
  Could not handshake: Error in the pull function. [IP: 185.xxx.xxx.xxx 443]
Err:2 https://nvidia.github.io/libnvidia-container/stable/deb/amd64  nvidia-container-toolkit 1.15.0-1
  Could not handshake: Error in the pull function. [IP: 185.xxx.xxx.xxx 443]
Err:3 https://nvidia.github.io/libnvidia-container/stable/deb/amd64  nvidia-docker2 2.14.0-1
  Could not handshake: Error in the pull function. [IP: 185.xxx.xxx.xxx 443]
E: Failed to fetch https://nvidia.github.io/libnvidia-container/stable/deb/amd64/./nvidia-container-toolkit-base_1.15.0-1_amd64.deb  Could not handshake: Error in the pull function. [IP: 185.xxx.xxx.xxx 443]
E: Failed to fetch https://nvidia.github.io/libnvidia-container/stable/deb/amd64/./nvidia-container-toolkit_1.15.0-1_amd64.deb  Could not handshake: Error in the pull function. [IP: 185.xxx.xxx.xxx 443]
E: Failed to fetch https://nvidia.github.io/libnvidia-container/stable/deb/amd64/./nvidia-docker2_2.14.0-1_all.deb  Could not handshake: Error in the pull function. [IP: 185.xxx.xxx.xxx 443]
E: Unable to fetch some archives, maybe run apt-get update or try with --fix-missing?

解决方法：

报错含义呐就是上述三个包，没有获取到（因为大家熟知的网络原因）。我们通过给出的三个http链接，本地将三个包下载后，上传到服务器，然后本地安装：

sudo apt install -y ./nvidia-*

4. 配置 Docker 使用 NVIDIA 运行时

编辑 Docker 配置文件 /etc/docker/daemon.json，确保包含以下内容：

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

5. 重启 Docker 服务

在更改配置后，重启 Docker 服务：

sudo systemctl restart docker

重启完后就可以正常使用了