brainstorming
brainstorming
You MUST use this before any creative work - creating features, building components, adding functionality, modifying behavior, or planning anything new, in software or out of it (a talk, a business, a renovation).
スキルを探す
開発・技術 · SKILL.md
cortex-m
cortex-mDevelop, export, build, and test ExecuTorch Cortex-M/CMSIS-NN workflows with PT2E quantization, graph rewrites, dialect tests, and Arm toolchain prerequisites.
Independent GitHub snapshot · Automated classification · Original documentation governs
スキルの説明
GitHubname: cortex-m description: "Develop, export, build, and test ExecuTorch Cortex-M/CMSIS-NN workflows with PT2E quantization, graph rewrites, dialect tests, and Arm toolchain prerequisites." disable-model-invocation: true metadata: disco-role: operating license: NOASSERTION
cortex-m
Use this sub-skill for ExecuTorch Cortex-M, CMSIS-NN, Arm MCU, Corstone/FVP, bare-metal, or Cortex-M quantizer/pass-manager tasks.
What This Owns
- PT2E quantization with
CortexMQuantizer. - Graph rewrite flow using
CortexMPassManagerand Cortex-M custom ops. - Distinguishing pure Python/dialect graph tests from implementation tests that require Arm tools/FVP.
- Bare-metal runner planning and troubleshooting.
Fast Path
- Decide whether the user is asking for graph/export correctness or hardware implementation correctness.
- For graph/export, use API reference and plan dialect tests first.
- For implementation, confirm Arm toolchain/FVP and any license/EULA prerequisites before running commands.
- Use the planner:
python scripts/plan_cortex_m_test.py --mode dialect
python scripts/plan_cortex_m_test.py --mode implementation --model conv2d
Cross-Links
- Generic source build issues:
../setup-build/SKILL.md. - Backend choice and Arm Ethos-U/VGF alternatives:
../backend-selection/SKILL.md. - Export runtime sequence outside Cortex-M-specific passes:
../export-runtime/SKILL.md.