机器学习 Step By Step：Linux 配置 Nvidia Docker 以支持 Nvidia GPU

1. 安装 Nvidia 相关依赖

CentOS7

yum install -y nvidia-container-toolkit nvidia-container-runtime

2. 配置 dockerd 启动命令

修改 Systemd Service 配置文件

mkdir -p /etc/systemd/system/docker.service.d
tee /etc/systemd/system/docker.service.d/override.conf <<EOF
[Service]
ExecStart=
ExecStart=/usr/bin/dockerd --host=fd:// --add-runtime=nvidia=/usr/bin/nvidia-container-runtime
EOF
systemctl daemon-reload
systemctl restart docker

修改 Docker Daemon configuration file

tee /etc/docker/daemon.json <<EOF
{
    "runtimes": {
        "nvidia": {
            "path": "/usr/bin/nvidia-container-runtime",
            "runtimeArgs": []
        }
    }
}
EOF
pkill -SIGHUP dockerd

你也可以将 nvidia 配置为默认的 runtime，将下面的配置加入 /etc/docker/daemon.json:

"default-runtime": "nvidia"

或者执行命令：

dockerd --add-runtime=nvidia=/usr/bin/nvidia-container-runtime

重启 dockerd

systemctl restart docker

3. 测试

docker run --runtime=nvidia --rm nvidia/cuda nvidia-smi

4. 踩坑笔记

• 异常1：启动 docker 之后，运行脚本，出现open3d-python==0.3.0.0 依赖库异常 ImportError: libGL.so.1: cannot open shared object file: No such file or direc tory。

  安装 libgl1-mesa-glx 库：

  apt-get update && apt-get install libgl1-mesa-glx
  

  如果输出以下 is not signed 错误，则需要首先将 nvidia 的 apt-get 软件源移走，然后再更新

  Ign:10 https://developer.download.nvidia.cn/compute/machine-learning/repos/ubuntu1804/x86_64  Release.gpg
  Reading package lists... Done
  W: GPG error: https://developer.download.nvidia.cn/compute/machine-learning/repos/ubuntu1804/x86_64  Release: The following signatures were invalid: BADSIG
  F60F4B3D7FA2AF80 cudatools <cudatools@nvidia.com>
  E: The repository 'https://developer.download.nvidia.com/compute/machine-learning/repos/ubuntu1804/x86_64  Release' is not signed.
  N: Updating from such a repository can't be done securely, and is therefore disabled by default.
  N: See apt-secure(8) manpage for repository creation and user configuration details.
  

  执行：

  mv /etc/apt/sources.list.d/cu͚da.list /etc/apt/sources.list.d/nvidia-ml.list /tmp
  

5. 参考

1. https://github.com/conda-forge/pygridgen-feedstock/issues/10
2. https://askubuntu.com/questions/408638/how-can-i-install-libgl1-mesa-glx