Skip to content

CONTRIBUTING.md

Latest commit

 

History

History
139 lines (81 loc) · 4.33 KB

CONTRIBUTING.md

File metadata and controls

139 lines (81 loc) · 4.33 KB

Contributing to OpenMath

Thank you for your interest in contributing to this project! 🎉
We welcome contributions that improve training, evaluation, usability, and documentation.

Code of Conduct

Please read and follow our Code of Conduct: CODE_OF_CONDUCT.md. All contributors, maintainers and participants are expected to adhere to it.

This repository focuses on QLoRA (4-bit) fine-tuning of Qwen2.5-Math-1.5B on the GSM8K dataset and sharing LoRA adapter weights.

📌 Scope of Contributions

This repository focuses on:

• QLoRA fine-tuning (4-bit) for Qwen2.5-Math-1.5B

• Training and evaluating on the GSM8K dataset

• Sharing LoRA adapter weights (not full model weights)

Please keep contributions meaningful, focused, and well-documented.

🚫 Out of Scope

To keep the repository focused, the following contributions are out of scope:

  • Uploading or sharing full model weights
  • Use or inclusion of proprietary or non-redistributable datasets
  • Changes unrelated to QLoRA fine-tuning, evaluation, or documentation
  • Large binary artifacts or checkpoints committed to the repository

🚀 How to Contribute

Step-by-Step Workflow:-

1. Fork the repository and clone your fork locally to your GitHub account

  git clone https://github.com/your-username/repository-name.git
  cd repository-name

2. Create a new branch for your change:

git checkout -b feature/your-feature-name

3. Make your changes

4. Commit your work with a clear, descriptive message

5. Open a Pull Request and explain:

• What you changed

• Why the change is needed

• The scope of the change (docs / evaluation / training / other)

PRs with clear explanations and limited scope are reviewed faster.

🎓 OSCG 2026 Contributors

This project participates in Open Source Connect Global (OSCG) 2026.

OSGC contributors are expected to:

  • Follow the general contribution guidelines in this document
  • Work on issues labeled or approved for OSCG participation
  • Communicate progress clearly in issue threads and pull requests
  • Ensure all contributions are well-documented and reproducible

For substantial changes, please open an issue first to discuss the scope with the maintainers.

💡 Suggested Contribution Areas

Area What to Contribute Expected Outcome
Training / Evaluation Scripts Improve or refactor training and evaluation code Cleaner scripts, better logging, reproducible runs
Experiments & Results Run experiments and report settings and metrics Documented accuracy and comparisons
New Datasets Add or integrate datasets (e.g., SVAMP, ASDiv) Improved generalization and evaluation
Decoding Strategies Enhance decoding to reduce repetition or errors More consistent and accurate outputs
Demos / Visualizations Add lightweight demos or visual tools Easier model interaction (e.g., Streamlit)
Documentation Improve README, comments, or examples Better onboarding for contributors
Bug Fixes / Refactoring Fix bugs or clean existing code More stable and maintainable code

Pull Request Requirements

Every PR must include:

  • A clear summary of what changed
  • Why the change is useful
  • Scope confirmation (e.g., docs-only, evaluation-only, training-related)

If the PR includes code changes (training or evaluation):

  • Provide reproducibility steps (commands, configs, seeds if applicable)
  • Mention dataset splits and evaluation metrics used
  • Clearly state any new dependencies or configuration changes

PRs with well-scoped, well-explained, and reproducible changes are preferred.

Note :- Please open an issue and get approval before starting any major changes.

🔧 Training Contribution Requirements

All training or fine-tuning related contributions must:

  • Use QLoRA (4-bit) for parameter-efficient fine-tuning
  • Follow the official prompt format defined by this repository
  • Clearly document training configurations, datasets, and evaluation metrics

Training contributions that do not follow these requirements may be requested to revise or may be declined.

License

By contributing, you agree that your contributions will be licensed under the Apache License 2.0.