docker-datascience-cuda

"docker-datascience-cuda" is a simple/easy-to-use template for setting up a python data science environment using Docker & VScode.

Features

• GPU available in container

• Jupyter notebook with VScode

• Useful extensions & libraries are available

  • Ubuntu-22.04 packages

    • git
    • vim
    • powerline

  • VScode extensions

    • git extensions
    • code highlight & completion

  • libralies for Japanese Python users

    • Mecab (mecab-ipadic-neologd)
    • japanese_matplotlib

• workspace that can be shared with the host

Requirement

• CUDA>=11.8
• Docker>=20.10.16
• Docker-Compose>=2.6.0
• VScode>=1.73.1
• Dev Container>=0.264.0
• git

Installation

1. Upgrade your CUDA Toolkit >= 11.8 (or your GPU Driver)
2. Download Docker-Desktop and setup
3. Download VScode and setup
4. Download Remote Development in VScode
5. Download Git and setup

Usage

1. Clone this repository

git clone https://github.com/hayashizaki-yu/docker-datascience-cuda.git

2. Open this folder with VScode

   • Click "files" in the upper left corner of VScode screen
   • Click "Open folder"
   • Select this folder

3. Open Container with Dev container

   • Click "><" icon in the lower left corner of VScode screen
   • Select "Reopen in container"
   • Initially it takes a while to start up

4. restart VScode

   • Initially, VScode extention for python will not be installed
   • You can resolve this problem by restarting VScode

Extra: install pytorch or tensorflow

• pytorch
  • run the command in container

pip install torch torch vision torchaudio

• tensorflow
  • run the command in container

pip install tensorflow-gpu

License

"docker-datascience-cuda" is under MIT license.