OpenParable
Skip to content
← Back to skills

Fla Mr Readiness

Development
fla-mr-readiness

Fla Mr Readiness is an Agent Skill for development published in the fla-org/flash-linear-attention repository on GitHub: a SKILL.md instruction file an AI agent can load to do this task.

Checklist and workflow for preparing an MR/PR in the FLA repo. Covers CONTRIBUTING.md compliance, test plan, benchmark evidence, and PR body structure.

View on GitHub ↗
QUICK START

How to use this skill

Bring this guide into your coding agent with a prompt tailored to the tool you use.

  1. Open your project in Codex.
  2. Copy and paste the prompt below.
  3. Review files and risks before approving any action.
Prompt to paste
Inspect the Agent Skill at https://github.com/fla-org/flash-linear-attention/blob/HEAD/.agents/skills/fla-mr-readiness/SKILL.md and all supporting files. Report provenance, permissions, scripts, dependencies, and security risks. Propose an exact installation into .agents/skills/fla-mr-readiness, but wait for my approval before writing files or running scripts.

Copying this prompt does not install or run the skill. Review third-party files before use.

FLA MR Readiness Skill

Use this skill before opening a pull request to make sure the change is well-scoped, well-tested, and well-documented.

Pre-flight checklist

  1. Read CONTRIBUTING.md

    • Confirm code style, docstring format, commit message conventions.
    • Make sure your branch is up to date with main (or the target branch).
  2. Confirm change scope

    • List the files you modified.
    • If the change spans multiple layers (kernel + model + benchmark script), note the dependency chain in the PR description.
  3. Check for duplicate work

    • Search open issues and PRs:
      gh pr list --repo fla-org/flash-linear-attention --state open --search "<keywords>"
      
    • If a related PR exists, comment on it rather than opening a competing one.
  4. Run dependent tests

    • Find tests affected by your change:
      python scripts/find_dependent_tests.py <changed_file_or_dir>
      
    • Run those tests locally and ensure they pass.
    • If a test is flaky, retry once; if it still fails, explain why in the PR.
  5. Performance evidence (if touching kernel code)

    • See fla-nvidia-performance skill for the full evidence requirements.
    • At minimum: before/after benchmark on the same hardware, dense + varlen workloads if applicable, and a summary of any NCU profiling you did.
  6. Write PR summary

    • Use the structure below.
  7. Code style review

    • Follow CONTRIBUTING.md for Python style, docstrings, comments, and commit prefixes.
    • In tests and public code, use device/platform wrappers from fla.utils (device, device_platform, IS_NVIDIA, IS_NVIDIA_HOPPER, IS_NVIDIA_BLACKWELL, IS_AMD, IS_INTEL) instead of new direct torch.cuda platform checks. Add a small fla.utils helper first when the existing wrappers are not enough.
    • Keep NVIDIA-only profiling commands in performance docs or scripts, not in generic correctness tests.

Suggested PR body structure

## Summary
One-paragraph description of what changed and why.

## Test plan
- Unit tests added/modified: `<list>`
- Dependent tests run: `<list>`
- Varlen / CP / model tests: `<yes/no + details>`

## Benchmark / NCU
- Hardware: `<e.g., H100>`
- Workload: `<batch, seq_len, dtype>`
- Before: `<throughput or latency>`
- After: `<throughput or latency>`
- Conclusion: `<improvement / neutral / trade-off>`

## Breaking changes
- None / list any API or behavior changes.

Important reminders

  • Do not put raw performance numbers without context. Always include:

    • workload shape (batch, seq_len, heads, dims, dtype)
    • hardware model
    • benchmark command used
    • before vs after
    • your conclusion
  • Do not commit .ncu-rep files or raw profile dumps. Summarize results in the PR body and keep artifacts local.

  • No busywork PRs: bundle trivial cleanups into a substantive change; do not open a PR for a single typo unless it is part of a larger fix.