- Deploy reward model:
sky launch serve_rm/serve_rm.yaml
- Jupyter notebook:
jupyter notebook best_hn/generate_titles/prepare_data.ipynb
- Run Python module:
python -m best_hn.generate_titles.<module_name>
- Run a single test:
pytest -xvs test_file.py::test_function_name
- Imports: stdlib first, third-party packages second, local modules last
- Naming: snake_case for variables/functions, PascalCase for classes
- Type hints: Use explicit type annotations for function parameters and returns
- Error handling: Use try/except with specific exceptions, avoid bare except
- Models: Use Pydantic for request/response models
- Docstrings: Google style docstrings for classes and functions
- Line length: 88 characters (Black default)
best_hn/generate_titles
: Main package for HN title generationserve_rm
: FastAPI server for reward model inference