"We demand rigidly defined areas of doubt and uncertainty."
Generic template to bootstrap your PyTorch project, read more in the documentation.
If you already know cookiecutter, just generate your project with:
cookiecutter https://github.com/grok-ai/nn-templateOtherwise
Cookiecutter manages the setup stages and delivers to you a personalized ready to run project.Install it with:
pip install cookiecutter
More details in the documentation.
- Actually works for research!
- Guided setup to customize project bootstrapping;
- Fast prototyping of new ideas, no need to build a new code base from scratch;
- Less boilerplate with no impact on the learning curve (as long as you know the integrated tools);
- Ensure experiments reproducibility;
- Automatize via GitHub actions: testing, stylish documentation deploy, PyPi upload;
- Enforce Python best practices;
- Many more in the documentation;
Avoid writing boilerplate code to integrate:
- PyTorch Lightning, lightweight PyTorch wrapper for high-performance AI research.
- Hydra, a framework for elegantly configuring complex applications.
- Hugging Face Datasets,a library for easily accessing and sharing datasets.
- Weights and Biases, organize and analyze machine learning experiments. (educational account available)
- Streamlit, turns data scripts into shareable web apps in minutes.
- MkDocs and Material for MkDocs, a fast, simple and downright gorgeous static site generator.
- DVC, track large files, directories, or ML models. Think "Git for data".
- GitHub Actions, to run the tests, publish the documentation and to PyPI automatically.
- Python best practices for developing and publishing research projects.