cookiecutter-docker-jupyterlab

A cookiecutter template to create Docker environment for ML/DL tasks.

The resulting environment enables users to:

• install Python libraries via pip
• use docker-compose to allow running multiple containers in one go
• use JupyterLab and TensorBoard from their local web browser
• use Commuter to browse notebooks

Related Work

• cookiecutter-docker-science
• cookiecutter-dl-work

Usage

Options

• author_name: your name
• project_name: your project name
• project_description: description of your project (this will be inserted to README.md)
• use_nvidia_docker: whether to use nvidia-docker or not
• formatter_type: Python formatter
• jupyter_port: a port to access JupyterLab running on a container
• tensorboard_port: a port to access TensorBoard running on a container
• commuter_port: a port to access commuter running on a container

Generate project and launch Docker container

Run following commands:

$ cookiecutter https://github.com/kazuhirokomoda/cookiecutter-docker-jupyterlab
# cd to generated project dir
$ docker-compose build # (or docker-compose build --no-cache)
$ docker-compose up -d

Access following URLs from web browser:

• (JupyterLab) http://localhost:<jupyter_port>
• (TensorBoard) http://localhost:<tensorboard_port>
• (commuter) http://localhost:<commuter_port>