My computer env.
Ubuntu 16.04 LTS, Nvidia GTX 1080ti
Note
The latest graphic driver should be installed in advance
1. install docker
(https://docs.docker.com/install/linux/docker-ce/ubuntu/#install-docker-ce-1)
a. list the version
$ sudo apt-cache madison docker-ce
b. install specific version
$ sudo apt-get install docker-ce=18.03.1~ce-0-ubuntu
If nothing works, follows
https://unix.stackexchange.com/questions/363048/unable-to-locate-package-docker-ce-on-a-64bit-ubuntu
2. install nvidia-docker (the latest graphic driver should be installed in advance)
(http://www.kwangsiklee.com/ko/2018/02/%EC%A0%95%EB%A6%AC-ubuntu%EC%84%9C%EB%B2%84%EC%97%90-nvidia-docker-%EC%84%A4%EC%B9%98-%ED%95%98%EA%B8%B0/)
a. add repository
$ curl -s -L https://nvidia.github.io/nvidia-docker/gpgkey | \
sudo apt-key add -
$ curl -s -L https://nvidia.github.io/nvidia-docker/ubuntu16.04/amd64/nvidia-docker.list | \
sudo tee /etc/apt/sources.list.d/nvidia-docker.list
$ sudo apt-get update
b. install package
$ sudo apt-get install -y nvidia-docker2
c. shut down
$ sudo pkill -SIGHUP dockerd
3. download and run tensorflow latest docker (gpu version)
.this command will download the latest tensorflow with all required packages
.after the download, a contrainer will be running
a) download and run
$ nvidia-docker run -it -p 8888:8888 tensorflow/tensorflow:latest-devel-gpu
b) check
$python
$import tensorflow as tf
$print(tf.__version__)
c) exit
4. create a contrainer
a) check image name
$ nvidia-docker images
b) create a container
$ nvidia-docker run -it -d -p 8888:8888 --name container_name -v /home/dooseop/tmp:/tmp image_name
c) excecute container
$ nvidia-docker exec -it container_name /bin/bash