Hello Docker

I decided to use Docker to manage my development environments (I use Ubuntu 20.04.02 LTS.).

Installing Docker on Ubuntu

• Type the following commands in your terminal to install Docker.

1. Set up the repository

$ sudo apt-get update

$ sudo apt-get install -y apt-transport-https
$ sudo apt-get install -y ca-certificates
$ sudo apt-get install -y curl
$ sudo apt-get install -y gnupg-agent
$ sudo apt-get install -y software-properties-common

$ curl -fsSL https://download.docker.com/linux/ubuntu/gpg | sudo apt-key add -
$ sudo apt-key fingerprint 0EBFCD88
$ sudo add-apt-repository "deb [arch=amd64] https://download.docker.com/linux/ubuntu $(lsb_release -cs) stable"
$ sudo apt-get update
$ sudo apt-get install docker-ce docker-ce-cli containerd.io

2. Install docker engine -community

$ sudo apt-get update
$ sudo apt-get install docker-ce docker-ce-cli containerd.io

3. usermod

• To run docker without typing sudo,

$ sudo usermode -aG docker $USER

• This requires reboot of your system.

Installing Nvidia Docker

• Nvidia Docker should be installed to use a GPU (infrastructure) in your system, because Docker is separated with the OS.
• Nvidia driver need to be installed on your system prior to install Nvidia Docker. This can be checked at Settings > Software & Updates.

Settings > Software & Updates

• Nvidia-Docker can be installed using the following commands

$ 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

$ sudo apt-get update & sudo apt-get install -y nvidia-container-toolkit
$ sudo systemctl restart docker

Building a Pytorch container

1. Downloading an image

• Pytorch containers, built by many users, can be pulled to your system.
• Docker containers are uploaded on docker hub.
• In my case, I pulled a pytorch with pytorch-geometric container.

$ docker pull fjvallarino/pytorch-geometric

• The following command lists the downloaded images

$ docker images

2. Building a Pytorch container and utilizing Nvidia-docker

• The first step is making a share directory, e.g.,

$ cd ~
$ mkdir share

• Now we define a container whose name is pytorch and employ the downloaded image fjvallarino/pytorch-geometric.
  • ~/share: a file sharable space between my OS and the container
  • --gpus all: utilized GPUs via Nvidia docker

$ docker run -itd --name pytorch fjvallarino/pytorch-geometric -v ~/share:/root/share -p 8888:8888 --gpus all

• You can list the container

$ docker ps -a

• A container can be removed

$ docker rm [CONTAINER ID]

• You can run the container pytorch

$ docker start pytorch
$ docker exec -it pytorch bash

• Check GPUs running

root@...:/workspace# nvidia-smi

• Run the python and test Pytorch

root@...:/workspace# python3

import torch
torch.cuda.is_available()
x = torch.rand(5.3)

• Checking the file share folder
  • After running a container, current path is workspace
  • The defined share folder path is

root@ ... :/worksapce# cd ..
root@ ... :/worksapce# cd root
root@ ... :/worksapce# ls
share

• Stop the pytorch container

docker container stop pytorch

3. Install Jupyter Notebook on a container

I am very happy to say docker supports Anaconda and we don’t need to install on a container. And jupyter notebook can be installed.

conda install jupyter

And start jupyter notebook

jupyter notebook --ip=0.0.0.0 --port=8888 --allow-root

Other approach is using VSCODE

References

• Pytorch and Jupyternotebook on Docker (in Korean)
• Getting started Docker (in Korean)

Future plan

• Upload my IMGs