python_ml_template

First actions:

When setting up the repo (and potentially also the venv) from scratch, a few one-time actions are needed:

• install pip dependencies
• pre-commit install --install-hooks -t pre-commit -t commit-msg
• git config branch.master.mergeOptions "--squash
• Activate gh-pages web (otherwise the CI release will error here).

Features:

The following things are already integrated via CI, but can be run manually:

• Utest: python -m unittest

• utest with py coverage: coverage run -m unittest

• flake8

• all

• using pre-commit, added commitizen to pre-commit (remember to pre-commit install --install-hooks -t pre-commit -t commit-msg). This enforces "conventional commits" style: https://www.conventionalcommits.org/en/v1.0.0/#summary To commit, reecommended to pip install commitizen and then commit using: cz c (or cz c --retry if the last one failed).

• docs from scratch:

  1. Add docs folder and requirements.txt with sphinx and sphinx-rtd-theme

• Centralized version and metadata. Setup works with very few parameters

• To enforce squash merging to master, issue git config branch.master.mergeOptions "--squash" (info: https://stackoverflow.com/a/37828622)

• GH pages action. Make sure that the repo server has publishing enabled, otherwise it will error.

• PyPI: need a regular and a test account. Create a token for GH actions (if global only need to do this once). Then, in the GH repo, add that token under secrets->pypi. https://pypi.org/manage/account/token/

Further feature ideas/TODOs:

• TUT init check_installation
• autoimports in each init
• dcase test tools
• pylintrc?
• custom "asteroid" sphinx theme
• mypy (not needed in research code)
• gitignore.io
• dependabot https://dependabot.com/github-actions/
• Ignore commits in changelog: https://github.com/conventional-changelog/conventional-changelog/issues/342
• Deploy to conda-forge
• change docstring style to napoleon
• add doctest
• Add changelog to autodocs
• Integrate wiki. Add wiki to autodocs
• Autocomment: https://github.com/marketplace/actions/label-commenter

TODO:

2. CML+ Complete ML project
3. Generalize runner to GPU, home and GitLab