Skip to content

CONTRIBUTING.md

Latest commit

 

History

History
57 lines (43 loc) · 2.3 KB

CONTRIBUTING.md

File metadata and controls

57 lines (43 loc) · 2.3 KB
Raw

Contributing to dNATY

Thanks for considering a contribution. This document covers the workflow and the project's conventions.

Getting started

  1. Fork the repository
  2. Clone your fork: git clone https://github.com/YOUR-USERNAME/dNaty.git
  3. Create a branch: git checkout -b feature/your-feature
  4. Make changes and commit with clear messages
  5. Push to your fork and open a Pull Request

Project layout

dNaty/
├── dnaty/
│   ├── compress.py              # public API: compress, compress_cnn,
│   │                            #   compress_with_backbone, prune_conv_channels
│   ├── result.py                # CompressResult + load()
│   ├── evolution/evolver.py     # DnatyEvolver / CnnEvolver (NSGA-II)
│   ├── core/                    # DynamicMLP, DynamicCNN, Individual, episodic memory
│   ├── operators/               # structural mutation operators
│   ├── training/local_train.py  # fast local trainer
│   ├── monitoring/              # DriftDetector, ProductionTracker
│   └── utils/flops_counter.py   # count_flops, flops_by_layer
├── scripts/                     # benchmarks + stress_adversarial.py
└── tests/                       # pytest suite

Conventions

  • Python: type hints, docstrings on public functions, Black-compatible formatting (88 cols)
  • Commits: feat:, fix:, docs:, test:, refactor:, perf:
  • No fabricated numbers: every benchmark figure must come from a script in scripts/ that reproduces it

Testing

pytest tests/ -q                          # full local suite
pytest tests/test_sanity.py -q            # fast smoke (what CI runs first)
python scripts/stress_adversarial.py      # 38 adversarial scenarios

CI runs test_sanity, test_edge_cases, test_reproducibility and test_benchmark_market on Python 3.10/3.11. Heavy NAS suites run locally. A PR should keep the full local suite green.

Reporting issues

Use the GitHub issue templates:

  • Bug report — what happened, steps to reproduce, expected result, dnaty.__version__ and torch version
  • Feature request — the idea, the use case, why it belongs in scope (CPU-only NAS for MLP/CNN-head compression)

Code of conduct

Be respectful. Zero tolerance for harassment.