1.安装Docker 19.03.5

sudo apt-get update
sudo apt-get -y install apt-transport-https ca-certificates curl software-properties-common

curl -fsSL http://mirrors.aliyun.com/docker-ce/linux/ubuntu/gpg | sudo apt-key add -
sudo add-apt-repository "deb [arch=amd64] http://mirrors.aliyun.com/docker-ce/linux/ubuntu $(lsb_release -cs) stable"
sudo apt-get -y update
apt-cache madison docker-ce
sudo apt-get -y install docker-ce=5:19.03.5~3-0~ubuntu-bionic
sudo docker version


mkdir -p /etc/docker
cat <<EOF > /etc/docker/daemon.json 
{
    "hosts": [ "unix:///var/run/docker.sock" ],
    "mtu": 1454,
    "debug": false,
    "experimental": true,
    "metrics-addr": "0.0.0.0:1337",
    "selinux-enabled": false,
    "registry-mirrors": [ "http://hub-mirror.c.163.com","https://2508xbaf.mirror.aliyuncs.com"],
    "insecure-registries":[
        "gcr.io",
        "quay.io",
        "registry.cn-hangzhou.aliyuncs.com"
    ],
    "exec-opts": [ "native.cgroupdriver=cgroupfs" ],
    "exec-root": "/var/run/docker",
    "data-root": "/data/docker/graph",
    "storage-driver": "overlay2",
    "storage-opts": [ "overlay2.override_kernel_check=true" ],
    "live-restore": true,
    "max-concurrent-downloads": 10,
    "max-concurrent-uploads": 5,
    "shutdown-timeout": 15,
    "oom-score-adjust": -999,
    "default-shm-size": "64M",
    "log-driver": "json-file",
    "log-level": "warn",
    "log-opts": { "max-size": "5000m", "max-file": "10" },
    "default-address-pools":[{"base":"172.16.0.0/16","size":24}],
    "seccomp-profile": "",
    "no-new-privileges": false,
    "runtimes": {
        "nvidia": {
            "path": "nvidia-container-runtime",
            "runtimeArgs": []
        }
    }
}
EOF

systemctl daemon-reload
systemctl enable docker

2.安装nvidia-container-runtime

curl -s -L https://nvidia.github.io/nvidia-docker/gpgkey | sudo apt-key add -
distribution=$(. /etc/os-release;echo $ID$VERSION_ID)
echo $distribution
curl -s -L https://nvidia.github.io/nvidia-docker/$distribution/nvidia-docker.list | sudo tee /etc/apt/sources.list.d/nvidia-docker.list
wget https://nvidia.github.io/nvidia-container-runtime/ubuntu14.04/amd64/./nvidia-container-runtime-hook_1.4.0-1_amd64.deb
sudo apt-get -y update
sudo apt-get install -y nvidia-container-toolkit
sudo apt-get install -y nvidia-container-runtime
sudo systemctl restart docker

3.安装内核并重启

wget https://*kernel/linux-headers-4.15.0-041500-generic_4.15.0-041500.201802011154_amd64.deb
wget https://*/kernel/linux-headers-4.15.0-041500_4.15.0-041500.201802011154_all.deb
wget https://*/kernel/linux-image-4.15.0-041500-generic_4.15.0-041500.201802011154_amd64.deb

dpkg -i linux-headers-4.15.0-041500_4.15.0-041500.201802011154_all.deb
dpkg -i linux-headers-4.15.0-041500-generic_4.15.0-041500.201802011154_amd64.deb
dpkg -i linux-image-4.15.0-041500-generic_4.15.0-041500.201802011154_amd64.deb
sudo update-grub
reboot

4.安装nvidia驱动

wget http://us.download.nvidia.com/tesla/440.33.01/NVIDIA-Linux-x86_64-440.33.01.run
chmod +x NVIDIA-Linux-x86_64-440.33.01.run
./NVIDIA-Linux-x86_64-440.33.01.run

5.检测nvidia驱动

nvidia-smi  -pm 1

docker run --runtime=nvidia nvidia/cuda:9.0-base nvidia-smi
docker run --gpus all nvidia/cuda:9.0-base nvidia-smi