Benchmarks LLM inference and drives GPU kernel optimization with Magpie. Use when the user wants to benchmark vLLM, SGLang, or Atom; capture torch traces; post-process inference traces with TraceLens into prefill/decode and roofline reports; identify top bottleneck kernels or map profiler names to source; analyze or compare HIP, CUDA, PyTorch, or Triton kernels; validate and rank optimized…
安装
npx skills add amd/skills --skill magpie-kernel-evaluatornpx skills add amd/skills --skill magpie-kernel-evaluator -a claude-code -gnpx skills add amd/skills --skill magpie-kernel-evaluator -a codex -ggemini skills install https://github.com/amd/skills.git --path skills/magpie-kernel-evaluatorgh skill install amd/skills magpie-kernel-evaluatorgit clone --depth 1 https://github.com/amd/skills /tmp/amd__skills
mkdir -p ~/.claude/skills ~/.agents/skills
cp -R /tmp/amd__skills/skills/magpie-kernel-evaluator ~/.claude/skills/magpie-kernel-evaluator # Claude Code
cp -R /tmp/amd__skills/skills/magpie-kernel-evaluator ~/.agents/skills/magpie-kernel-evaluator # Codex, Cursor, Gemini CLI, Copilot, OpenCode