Skip to content

CONTRIBUTING.md

Latest commit

 

History

History
110 lines (75 loc) · 2.46 KB

CONTRIBUTING.md

File metadata and controls

110 lines (75 loc) · 2.46 KB

Contributing to wav2vec2.cpp

Thank you for your interest in contributing! This document provides guidelines for contributing to wav2vec2.cpp.

Code of Conduct

Be respectful and constructive in all interactions.

Getting Started

  1. Fork the repository
  2. Clone your fork: git clone https://github.com/YOUR_USERNAME/wav2vec2.cpp
  3. Create a branch: git checkout -b feature/your-feature
  4. Make your changes
  5. Test thoroughly
  6. Submit a pull request

Development Setup

# Build
mkdir build && cd build
cmake -DGGML_METAL=ON ..
make -j

# Test
./bin/wav2vec2-cli -m models/wav2vec2-phoneme/ggml-model-f16.bin -f samples/jfk.wav

Pull Request Guidelines

Before Submitting

  • Code compiles without warnings
  • Existing tests pass
  • New functionality includes tests
  • Accuracy metrics unchanged (run scripts/eval_pytorch.py)
  • Code follows project style (see CONVENTIONS.md)

PR Format

Title: <scope>: <description>

Examples:

  • wav2vec2: fix attention mask handling
  • examples: add streaming inference
  • docs: update build instructions

Description:

  • What does this PR do?
  • Why is this change needed?
  • How was it tested?

One PR Per Change

Submit separate PRs for unrelated changes. Don't mix bug fixes with new features.

Code Style

  • 4-space indentation
  • 120 character line limit
  • snake_case for functions and variables
  • wav2vec2_ prefix for public API functions
  • See docs/CONVENTIONS.md for details

AI-Generated Code

If you use AI tools (GitHub Copilot, ChatGPT, Claude, etc.) to generate code, please disclose this in your PR description. This project was itself "vibe coded" with AI assistance, so AI-assisted contributions are welcome with transparency.

Testing

Unit Tests

cd build && ctest

Accuracy Validation

Compare against PyTorch reference:

python scripts/eval_pytorch.py \
    --audio samples/jfk.wav \
    --model models/wav2vec2-phoneme/ggml-model-f16.bin

Expected: PER < 5% vs PyTorch, or improvement over current baseline.

Reporting Issues

Bug Reports

Include:

  • Platform (macOS/Linux/Windows, CPU/GPU)
  • Model used
  • Steps to reproduce
  • Expected vs actual behavior
  • Relevant logs

Feature Requests

Describe:

  • The problem you're trying to solve
  • Your proposed solution
  • Alternatives considered

License

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