Thank you for your interest in contributing to Autodistill!
We welcome any contributions to help us improve the quality of autodistill
and expand the range of supported models.
We welcome contributions to:
- Add a new base model (see more guidance below).
- Add a new target model (see more guidance below).
- Report bugs and issues in the project.
- Submit a request for a new task or feature.
- Improve our test coverage.
Autodistill is designed with modularity in mind. We want autodistill
to extend across different models and problem types, providing a consistent interface for distilling models.
We welcome contributions that add new models to the project. Before you begin, please make sure that another contributor has not already begun work on the model you want to add. You can check the project README for our roadmap on adding more models.
To add a new model, create a new repo that requires autodistill
and implement the BaseModel
or TargetModel
class for your task. You can use the existing models as a guide for how to structure your code.
Finally, you will need to add documentation for your model and link to it from the autodistill
README. You can add a new page to the docs/models
directory that describes your model and how to use it. You can use the existing model documentation as a guide for how to structure your documentation.
First, fork this repository to your own GitHub account. Create a new branch that describes your changes (i.e. line-counter-docs
). Push your changes to the branch on your fork and then submit a pull request to this repository.
When creating new functions, please ensure you have the following:
- Docstrings for the function and all parameters.
- Examples in the documentation for the function.
- Created an entry in our docs to autogenerate the documentation for the function.
All pull requests will be reviewed by the maintainers of the project. We will provide feedback and ask for changes if necessary.
PRs must pass all tests and linting requirements before they can be merged.
We provide two handy commands inside the Makefile
, namely:
make style
to format the codemake check_code_quality
to check code quality (PEP8 basically)
So far, there is no types checking with mypy. See issue.
pytests
is used to run our tests.