Skip to content

Contributing to UniLLM

Thank you for your interest in contributing to UniLLM! This guide will help you get started.

Ways to Contribute

Code Contributions

  • New Models - Implement support for additional model architectures
  • Bug Fixes - Fix issues in existing code
  • Performance - Optimize inference speed and memory usage
  • Features - Add new capabilities to the runtime

Non-Code Contributions

  • Documentation - Improve guides, examples, and API docs
  • Testing - Write tests, report bugs, verify fixes
  • Examples - Create sample applications and tutorials
  • Feedback - Share your experience and suggestions

Quick Start

  1. Fork and clone the repository
  2. Set up development environment
  3. Make changes following our guidelines
  4. Test your changes
  5. Submit a pull request
# Clone your fork
git clone https://github.com/YOUR_USERNAME/unillm.git
cd unillm

# Add upstream remote
git remote add upstream https://github.com/unillm/unillm.git

# Create a branch
git checkout -b feature/my-contribution

# Make changes, then commit
git add .
git commit -m "Add feature X"

# Push to your fork
git push origin feature/my-contribution

Contribution Areas

Priority Areas

Area Description Difficulty
GPU Backends CUDA and Metal acceleration High
KV Caching Efficient autoregressive generation Medium
New Models Add model architectures Medium
Quantized Inference Use quantized weights directly High
Production Server HTTP API implementation Medium

Good First Issues

Look for issues labeled good-first-issue:

  • Documentation improvements
  • Simple bug fixes
  • Test additions
  • Small feature enhancements

Code Guidelines

General Principles

  1. Follow the three-layer abstraction - TensorCore, ModelCore, WeightLoaderCore
  2. Use established patterns - Look at existing models for examples
  3. Write tests - All new code should have tests
  4. Document - Add comments for complex logic
  5. Keep it simple - Avoid unnecessary complexity

Rust Style

// Good: Clear, descriptive names
fn forward_attention(&self, hidden: &Tensor, mask: &Tensor) -> Result<Tensor>

// Bad: Cryptic abbreviations
fn fwd_attn(&self, h: &Tensor, m: &Tensor) -> Result<Tensor>

See Code Style Guide for details.

Pull Request Process

Before Submitting

  1. Sync with upstream 

    git fetch upstream
git rebase upstream/main

  2. Run tests 

    cargo test --lib -p runtime

  3. Check formatting 

    cargo fmt --check

  4. Run clippy 

    cargo clippy -- -D warnings

PR Guidelines

  • Clear title - Describe what the PR does
  • Description - Explain why and how
  • Small PRs - Easier to review
  • One thing - Each PR should do one thing
  • Tests - Include tests for new code

Review Process

  1. Automated checks - CI must pass
  2. Code review - At least one approval
  3. Testing - Manual testing if needed
  4. Merge - Squash and merge

Getting Help

Communication

  • Issues - For bugs and feature requests
  • Discussions - For questions and ideas
  • Pull Requests - For code contributions

Resources

Recognition

Contributors are recognized in:

  • CONTRIBUTORS.md file
  • Release notes
  • README acknowledgments

License

By contributing, you agree that your contributions will be licensed under the same license as the project (MIT/Apache-2.0).

Code of Conduct

We are committed to providing a welcoming and inclusive environment. Please:

  • Be respectful and constructive
  • Focus on the code, not the person
  • Help others learn and grow
  • Report unacceptable behavior

Thank you for contributing to UniLLM!