docs: add contributing guidelines for CogniXpert v1.0
Browse filesAdd CONTRIBUTING.md file with contribution guidelines, development setup instructions, pull request guidelines, coding style recommendations, and safety considerations for the CogniXpert project
- CONTRIBUTING.md +57 -0
CONTRIBUTING.md
ADDED
|
@@ -0,0 +1,57 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# Contributing to CogniXpert v1.0
|
| 2 |
+
|
| 3 |
+
Thanks for your interest in improving CogniXpert. Contributions of code, docs, evaluations, and safety improvements are welcome.
|
| 4 |
+
|
| 5 |
+
## Ways to Contribute
|
| 6 |
+
|
| 7 |
+
- Report bugs or issues
|
| 8 |
+
- Improve documentation and examples
|
| 9 |
+
- Propose new evaluation scripts or prompts
|
| 10 |
+
- Contribute training or alignment recipes
|
| 11 |
+
- Optimize inference and memory usage
|
| 12 |
+
|
| 13 |
+
## Development Setup
|
| 14 |
+
|
| 15 |
+
- Python 3.10+
|
| 16 |
+
- `pip install -U transformers unsloth peft bitsandbytes`
|
| 17 |
+
- Optional GPU acceleration: CUDA 12.x with a recent NVIDIA driver
|
| 18 |
+
|
| 19 |
+
## Pull Request Guidelines
|
| 20 |
+
|
| 21 |
+
- Fork the repo and create a topic branch
|
| 22 |
+
- Keep changes focused and incremental
|
| 23 |
+
- Update docs and examples when behavior changes
|
| 24 |
+
- Add usage notes for new configs or flags
|
| 25 |
+
- Ensure code is free of secrets or proprietary data
|
| 26 |
+
|
| 27 |
+
## Coding and Docs Style
|
| 28 |
+
|
| 29 |
+
- Prefer clear, simple Python samples
|
| 30 |
+
- Use `device_map="auto"` for examples unless reasoned otherwise
|
| 31 |
+
- Keep README snippets runnable
|
| 32 |
+
- Write concise commit messages in imperative mood
|
| 33 |
+
|
| 34 |
+
## Safety and Scope
|
| 35 |
+
|
| 36 |
+
- Do not claim medical capability; include help‑seeking guidance
|
| 37 |
+
- Avoid training data that identifies individuals
|
| 38 |
+
- Note limitations and potential biases in evaluations
|
| 39 |
+
|
| 40 |
+
## Issue Triage
|
| 41 |
+
|
| 42 |
+
- `bug`: malfunction or incorrect behavior
|
| 43 |
+
- `docs`: documentation improvements
|
| 44 |
+
- `perf`: performance or memory optimization
|
| 45 |
+
- `safety`: alignment or guardrails
|
| 46 |
+
- `feature`: new recipes or capabilities
|
| 47 |
+
|
| 48 |
+
## Release and Weights
|
| 49 |
+
|
| 50 |
+
- LoRA adapters should include `adapter_config.json` and weights files
|
| 51 |
+
- Reference base model IDs and versions used during training
|
| 52 |
+
- Document data sources and filtering where possible
|
| 53 |
+
|
| 54 |
+
## License
|
| 55 |
+
|
| 56 |
+
By contributing, you agree your contributions are licensed under AGPL‑3.0.
|
| 57 |
+
|