Docker-based Deep Learning Model Development Environment

This repository contains a Docker-based development environment for deep learning model development.

It can be easily used just by adding it as a submodule under the project folder of the target code.

Only one container is created and the name is automatically generated.

Prerequisites

Tested environment

• Ubuntu LTS (16.04 or 18.04)
• docker-ce (>= 18.09)
• docker-compose (>= 1.25.4) - This script will automatically upgrade docker-compose if necessary
• nvidia-docker2 (>= 2.0.3)

How to use

• Build docker image

  {PJT_DIR}$ ./docker/compose.sh -b

  -b : build an image
  -r : run a container
  -s : connect to shell(bash)
  -k : kill container (attach and kill)
  -d : down container (kill container and remove container, network and volumes)
  -t, --tensorboard : run tensorboard (default path: /home/${USER}/dev/checkpoints, port: 6006)
  -j, --jupyter : run jupyter notebook server (default port: 8888)
  --no-cache : build image from scratch(use no cache)

  This script makes ONLY 1 CONTAINER

• Attach the container using docker-compose

  {PJT_DIR}$ ./docker/compose.sh -s

• This script supports multiple argument

  {PJT_DIR}$ bash ./compose.sh -brs

  build -> run -> attach to shell (Automatically executed sequentially)

• To build specific service

  {PJT_DIR}$ bash ./compose.sh -bj

  build jupyter image -> run jupyter service (Automatically executed sequentially)

example)
ubuntu user ID: nounique
directory name : _GIT_REPO
location: _GIT_REPO/docker
buuild & run result:

$ docker ps
CONTAINER ID  IMAGE             COMMAND      NAMES
78487a99355e  gitrepo:nounique  "/bin/bash"  nounique_gitrepo_dev_1

Installation (Docker on host)

1. Remove default docker packages(old versions)

   sudo apt-get purge docker \
                      docker-engine \
                      docker.io \
                      lxc-docker 

2. Install required packages for installing docker

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

3. Import docker GPG key to verified packages signiture

   curl -fsSL https://download.docker.com/linux/ubuntu/gpg | sudo apt-key add -
   sudo apt-key fingerprint 0EBFCD88

4. Add docker repository

   sudo add-apt-repository \
        "deb [arch=amd64] https://download.docker.com/linux/ubuntu \
        $(lsb_release -cs) stable"
   sudo apt-get update

5. Install docker(docker-ce) and docker-compose

   sudo apt-get install docker-ce \
                        docker-ce-cli \
                        containerd.io \
                        docker-compose

   Give user a root permission

   sudo usermod -aG docker $USER

6. Install nvidia-docker2 Remove nvidia-docker1.0

   docker volume ls -q -f driver=nvidia-docker | \
   xargs -r -I{} -n1 docker ps -q -a -f volume={} | \
   xargs -r docker rm -f
   sudo apt-get purge nvidia-docker

   Add nvidia-docker repository

   curl -sL https://nvidia.github.io/nvidia-docker/gpgkey | sudo apt-key add -
   dist=$(. /etc/os-release;echo $ID$VERSION_ID)
   curl -sL https://nvidia.github.io/nvidia-docker/$dist/nvidia-docker.list | \
   sudo tee /etc/apt/sources.list.d/nvidia-docker.list
   sudo apt-get update

   Install nvidia-docker2

   sudo apt-get install nvidia-docker2
   sudo pkill -SIGHUP dockerd

   Set nvidia-docker as default-runtime of docker

   sudo vi /etc/docker/daemon.json

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

   sudo systemctl restart docker

   sudo chown "$USER":"$USER" /home/"$USER"/.docker -R
   sudo chmod g+rwx "/home/$USER/.docker" -R

   Re-login is required

   Test nvidia-docker

   docker run --rm -it nvidia/cuda nvidia-smi