Skip to content

CONTRIBUTING.md

Latest commit

 

History

History
45 lines (34 loc) · 1.65 KB

CONTRIBUTING.md

File metadata and controls

45 lines (34 loc) · 1.65 KB
Raw

Contributing

Thanks for your interest in GAE. This is a research code release; the notes below keep changes easy to review.

Setup

python -m venv .venv && source .venv/bin/activate
pip install -e .            # installs the `gae` facade and the src/ modules
pip install pytest ruff     # dev tools

pip install -e . makes both gae and the research packages (stage1, stage2, utils, metrics, video, data, cut3r_data, disc) importable, so you no longer need to set PYTHONPATH by hand.

Layout

  • gae/ — small public API. Prefer adding user-facing entry points here.
  • scripts/{demo,train,eval,data}/ — CLIs for generation, training, evaluation, and data prep.
  • src/ — research modules. See docs/CODEBASE.md for which files are load-bearing versus kept only for checkpoint compatibility.
  • configs/ — the released recipes. Keep config keys stable so published checkpoints keep loading.

Before you open a PR

ruff check .
python scripts/eval/validate_configs.py
python scripts/eval/smoke_test.py
pytest -q tests/test_configs.py tests/test_imports.py tests/test_prepare_data.py

The full training/eval path needs GPUs, DA3-GIANT, and packed datasets, so CI only runs the CPU-only config, import, and data tests above.

Conventions

  • Do not hard-code absolute machine paths. Read data roots from GAE_DATA_ROOT and model/cache locations from documented env vars.
  • Keep the public class names (GAECodec, GAEFlow) and paper equation references intact; map any renames in docs/METHOD.md.
  • New optional dependencies go in a pyproject.toml extra, imported lazily.