Skip to content

CONTRIBUTING.md

Latest commit

 

History

History
41 lines (30 loc) · 1.15 KB

CONTRIBUTING.md

File metadata and controls

41 lines (30 loc) · 1.15 KB
Raw

Contributing to mlx-flash

Thank you for your interest! This project targets Apple Silicon Macs running macOS 13+. Here is how to get started.

Development setup

# 1. Fork & clone
git clone https://github.com/YOUR_USERNAME/mlx-flash
cd mlx-flash

# 2. Create a virtual environment (Python 3.11+)
python -m venv .venv && source .venv/bin/activate

# 3. Install in editable mode with dev extras
pip install -e ".[dev]"

# 4. Run the test suite against a small 4B model
scripts/test_quick.sh

Pull-request guidelines

  • Target the main branch.
  • Keep each PR focused; one feature / fix per PR.
  • All public functions need docstrings and type annotations.
  • Run ruff check . && mypy mlx_flash before opening a PR.
  • Add or update tests for every behavioural change.
  • For Metal kernel changes, include before/after bench_flash.py numbers.

Architecture overview

See docs/architecture.md for a deep dive into the streaming pipeline.

Reporting bugs

Open a GitHub Issue with:

  1. macOS + chip version (e.g. macOS 14.5, M4 Air 16 GB)
  2. Model name & quantisation level
  3. Full traceback
  4. Output of python -c "import mlx; print(mlx.__version__)"