luau

Latent Unified Adaptive Upskilling. test.

A project to accelerate reinforcement learning in novel domains on novel tasks.

• Poetry
  • For dependency management, packaging, and publishing
• Ruff
  • For linting/formatting (it's FAST)
• GitHub Actions
  • For CI/CD
• Pytest
  • For testing
• pre-commit
  • For pre-commit hooks
• PyInvoke
  • For task running, because I hate make

To get started

1. Create a repository from a template.
2. Clone the new repo
3. If poetry isn't installed, you need to install it.
4. terminal cd into the project
5. run poetry install
6. Run poetry run inv setup

The setup will

• Setup the poetry environment (or use the existing one you're activated to)
• Install the dependencies
• Setup the pre-commit hooks
• Ask you for some project details (name, author, etc) and update the pyproject.toml file

Features

This sets up a basic set of checks to run. If you already have a virtual environment setup for this project, you can skip all the poetry run parts of the command as long as that environment is active. Example poetry run inv checks would be inv checks if the environment is active. I won't be putting poetry run in front of every command, but if you don't have a virtual environment setup, you'll need to run poetry run inv checks instead of inv checks.

Pre-commit

Pre-commit is used to run checks before you commit. You can run pre-commit run --all-files to run all the checks. The configuration for pre-commit is located in .pre-commit-config.yaml. If you find yourself unable to commit, this is almost certainly the reason. You need to install it for it to work on the client side. You can do this by running pre-commit install.

Ruff

Ruff is used for linting and formatting. You can run ruff check --fix to check and fix the code. ruff format will only format the code. The configuration for ruff is located in ruff.toml.

Pytest

Pytest is used for testing. You can run pytest to run all the tests. The CI is configured to run pytest -m "not skipci" so any test marked with @pytest.mark.skipci will not run in the CI pipeline. See main_test.py for an example of how to use this.

PyInvoke

PyInvoke is used for task running, and chosen because make is black magic to me. You can run inv --list to see all the available tasks. The tasks are located in the tasks folder. The checks task will run all the checks.

• inv --list will show you all the available tasks
• inv checks will run all the checks located in checks.py

GitHub

For pull requests, the pipeline will run inv checks and run all the formatting checks. It will run the all the pytests, let you know what fails and succeeds in the pull request itself as well as give you a code coverage report. The pipeline is located in .github/workflows/ci.yml.

All pytests marked @pytest.mark.skipci will not run in the pipeline. This is useful for tests that are slow, or require a specific environment to run. You can run these tests locally, but they will not run in the pipeline. You can see an example of this in main_test.py.

In order to get true coverage numbers in your report, the checks look for files in the src folder with a matching _test.py file in the tests folder. If it doesn't have one, it creates a skeleton to just import.
For example, main.py has a matching main_test.py file.

Theres also issue templates and rulesets for the repository.

Contributing

If you have any suggestions, please open an issue. If you'd like to contribute, please open a pull request. I'm always looking for ways to improve this template. I'm open to suggestions, but I'm also very opinionated. I'm trying to keep it as simple as possible while remaining good enough for production code.

Updating from template

If you want to update your project from the template, or add the template to an existing project. There's a handy inv task. Just run inv setup.update-from-template.

or you can do it manually with the following commands

git remote add template https://github.com/lite-dsa/python-template.git
git fetch template
git merge template/main --allow-unrelated-histories

Packages

pip install stable-baselines3[extra]
pip install swig
pip install gymnasium[box2d]
pip install minigrid
pip install ffio
pip install wandb
pip install scikit-image
pip install h5py
pip install seaborn

Making a video from image frames

ffmpeg -framerate 25 -i frame_%06d.png -c:v libx264 -profile:v high -crf 20 -pix_fmt yuv420p ../output.mp4