AGENT SKILL

repository-development

Edit RF-DETR source while following project contribution, test, style, docs, packaging, model-selection, and CI conventions.

License 许可
Apache-2.0
最近更新
2026-09-03
Stars
★ 351
Forks
27

01 前言 Front Matter

name
repository-development
license
Apache 2.0
metadata
{ "disco-role": "operating" }
description
Edit RF-DETR source while following project contribution, test, style, docs, packaging, model-selection, and CI conventions.
disable-model-invocation
true

02 资源文件 4

03 SKILL.md 正文

RF-DETR Repository Development

Use this sub-skill when you are changing RF-DETR source, tests, docs, packaging, or CI. It is for contributor workflows, not for end-user inference, training, or export recipes.

Start Here

  1. Read Contributor guidance before editing source, tests, docs, package metadata, or workflow files.
  2. Use Test selection to choose focused checks during development and full gates before handoff.
  3. If a check fails or optional backends are involved, use Troubleshooting.
  4. To get a dry recommendation from changed paths, run:
python sub-skills/repository-development/scripts/select_checks.py PATH [PATH ...]

The script prints commands only; it does not run tests or modify files.

Routing

Non-Negotiables

  • Use TDD: failing test first for bug fixes; comprehensive tests for features.
  • Every new or changed function/class needs type hints and a Google-style docstring; do not duplicate types inside docstrings.
  • Python files need the RF-DETR license header.
  • Run pre-commit run --all-files before any commit/handoff.
  • Default examples/docs/tests to RFDETRSmall / "rfdetr-small"; do not add new RFDETRBase, "rfdetr-base", segmentation preview, or detection preview usage. Keypoint preview is the only preview variant to use for new material.
  • Keep imports direct and module-scope unless there is a clear optional-dependency, circular-import, import-behavior-test, or startup-side-effect reason.
  • Prefer small visible duplication over a helper that adds indirection without reducing cognitive load.

04 同仓库更多 Skills

VectorSpaceLab
VectorSpaceLab/AREX-Skill

passnet-triton-opt

Make a single PassNet Triton kernel fast AND numerically faithful: performance model (when a replacement can win at all), block/grid/warp tuning, autotune policy, launch overhead, and per-op numeric recipes to pass the dtype baseline tolerances. Use when a pass matches and is correct but speedup below expected, or when correctness fails by small numeric margins.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

passnet-solve

END-TO-END playbook for solving one PassNet sample: analyze the computation graphs, decide the optimization strategy (what to fuse, what to replace, what to leave alone), drive the iteration loop, and maximize the sample score. This is the ENTRY skill — invoke it first for any PassNet optimization task; it tells you when to use passnet-pattern-fusion, passnet-triton-opt and passnet-feedback.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

passnet-skill

PassNet GPU kernel optimization via compiler passes. Design and implement Triton-based optimization passes, create pass files under ./pass_dir/, self-evaluate with pass_evaluator, and iterate to maximize GPU speedup.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

passnet-pattern-fusion

Author PassNet pass files: write patterns that actually MATCH the FX graph, pick fusion regions (which ops to absorb into one Triton kernel), and structure multi-pass submissions with the mandatory shared-dispatch architecture. Use when creating or fixing pattern()/replacement_args()/replacement_func() files, when a pass "failed to match", or when deciding how to fuse multiple kernels/ops.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

passnet-orchestrate

Round/batch planning and decision support for PassNet graph optimization. Use for multi-sample triage, worker allocation, eval-budget policy, and post-eval keep/revert decisions. For solving one concrete sample end to end, use passnet-solve as the entry skill; consult this skill only when you need broader planning or a specific decision gate.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

passnet-feedback

Fast, deterministic feedback for PassNet work: pre-flight pattern/match verification WITHOUT burning a GPU evaluation, per-node bottleneck analysis of a sample's graphs, and eval-log parsing into per-variant status + estimated score + failure classification. Use BEFORE every GPU evaluation (check_pattern), at the START of a sample (analyze_graph), and AFTER every evaluation (parse_eval_log).

★ 351