Skip to content

CONTRIBUTING.md

Latest commit

 

History

History
83 lines (57 loc) · 2.08 KB

CONTRIBUTING.md

File metadata and controls

83 lines (57 loc) · 2.08 KB

Contributing

This project uses OpenSpec. All meaningful changes start from specs and tracked changes under openspec/.

Start here

  1. Read the relevant capability spec in openspec/specs/.
  2. Check active changes in openspec/changes/.
  3. If your work changes behavior or scope, create or update a change proposal before editing code.
  4. Keep the branch focused and short-lived.
  5. Run validation and a review pass before merge.

For AI-specific guidance, see AGENTS.md.

Local setup

git clone https://github.com/LessUp/llm-speed.git
cd llm-speed

python3 -m venv .venv
. .venv/bin/activate

pip install -U pip setuptools wheel
pip install -r requirements.txt pytest hypothesis ruff pre-commit

To build the CUDA extension locally:

pip install -e .

Validation

ruff check cuda_llm_ops/ tests/ benchmarks/
pytest tests/ -v -m "not cuda"
pre-commit run --all-files

GPU-specific validation can be run separately with:

pytest tests/ -v -m cuda

Workflow

OpenSpec commands

/opsx:propose <change-name>
/opsx:apply <change-name>
/opsx:archive <change-name>

Branch and review discipline

  • Keep one branch per change or cleanup slice.
  • Merge promptly after a coherent slice; avoid stale local/cloud branch drift.
  • Use /review or an equivalent review pass before merge.
  • Prefer concise, project-specific docs and config over generic process sprawl.

Commit style

Use conventional commits such as:

fix(ci): simplify cpu-safe validation
docs(readme): align setup with OpenSpec workflow

Tooling

  • C/CUDA: clangd + cmake --preset default
  • Python: pyright/basedpyright + pyrightconfig.json
  • Hooks: pre-commit with Ruff, clang-format, and file hygiene
  • MCP: optional; prefer gh, OpenSpec commands, and targeted subagents unless a heavier integration clearly pays off

Current closeout scope

The repository is being normalized for final stabilization. bf16-support and flashattention-backward remain deferred backlog and are not part of the active closeout path.