Thanks for helping improve this project.
- Fork and clone the repo.
- Create the environment:
make env conda activate sagemaker-llm-optimizer
- Run checks before opening a PR:
make lint make test
- Use focused branches (
feat/...,fix/...,docs/...,chore/...). - Keep commits small and descriptive.
- Prefer conventional commit style (e.g.,
fix(ci): ...,docs: ...).
Please include:
- What changed and why
- Any config/runtime implications
- Validation evidence (lint/test output, benchmark dry-run output, screenshots for dashboard changes)
If your PR changes benchmark logic, config schema, or metrics calculations:
- Add/update tests where possible
- Mention backward-compatibility impact
- Call out any expected changes in result interpretation
- Ruff + mypy + pytest are required to pass in CI.
- Prefer explicit typing and clear comments around AWS/SageMaker side effects.
- Never commit real secrets, tokens,
.env, or account-specific identifiers. - Use placeholders in docs and examples.
- See
SECURITY.mdfor vulnerability reporting.