Docker environments for machine learning programs

• My basic settings to run machine learning programs on docker container

Environment

• Host machine

  • Ubuntu18.04
  • Nvidia driver: 440.100
  • CUDA Version: 10.2
  • Docker: 19.03.6, build 369ce74a3c
  • docker-compose: version 1.25.0, build 0a186604

• Dockerfiles

  • env-tf: for tensorflow2.3-gpu, tensorboard(port 6006)

  (I will add other settings in advance)

Run

• Set an environment variable

  > export UID

  • I wrote UID into ~/.bashrc.

• Run docker-compose

  # Run docker containers
  > docker-compose up -d
  
  # Enter to the docker container
  > exec -it container_name bash 

Warning: In docker container, you need to run command su current_user_name before making some files. Otherwise, the owner of them will be root.