Optimizes transformer attention with Flash Attention for 2-4x speedup and 10-20x memory reduction. Use when training/running transformers with long sequences (>512 tokens), encountering GPU memory issues with attention, or need faster inference. Supports PyTorch native SDPA, flash-attn library, H100 FP8, and sliding window attention.
安装
npx skills add NousResearch/hermes-agent --skill flash-attentionnpx skills add NousResearch/hermes-agent --skill flash-attention -a claude-code -gnpx skills add NousResearch/hermes-agent --skill flash-attention -a codex -ggemini skills install https://github.com/NousResearch/hermes-agent.git --path optional-skills/mlops/flash-attentiongh skill install NousResearch/hermes-agent flash-attentiongit clone --depth 1 https://github.com/NousResearch/hermes-agent /tmp/NousResearch__hermes-agent
mkdir -p ~/.claude/skills ~/.agents/skills
cp -R /tmp/NousResearch__hermes-agent/optional-skills/mlops/flash-attention ~/.claude/skills/flash-attention # Claude Code
cp -R /tmp/NousResearch__hermes-agent/optional-skills/mlops/flash-attention ~/.agents/skills/flash-attention # Codex, Cursor, Gemini CLI, Copilot, OpenCode