Last Updated: 2021-05-11

Protocol for Public Repositories

There are 3 kinds of repositories we host:

private : Only visible to other lab members. Useful for sharing quick research code or incomplete ideas.

public experimental : Minimal guarantees. Likely corresponds to a publication and might not be maintained moving forward.

public stable : Guaranteed to adhere to highest repo standards. Tests with good coverage. Merges accepted through PRs that pass CI. Good target to fork.

Repo Requirements

🌱 private

  • no requirements.

🌿 public experimental 🧪

  • Tidy code on main/master. No extraneous files that play no role in the functionality
  • README.md with installation/usage instructions
  • One working, documented example

🌲 public stable

  • Concise, tidy code
  • Good docstring coverage
  • README.md with installation/usage instructions
  • Working, documented examples
  • Packaging via setup.py or poetry
  • CI (Github Actions) on main/master branch
    • Formatting via black
    • Linting via flake8
    • Testing via pytest
    • >85% test coverage

See repo-template for a starting point.

General Requirements

  • name your repo with kebab-case: robot-learning
  • name your package with snake_case: robot_learning
  • use tags to indicate stable or experimental in public repositories
  • use other informative tags like rl or manipulation when appropriate