Add EAGLE-3 or draft-model speculative decoding to a Jetson vLLM server when TPOT is the bottleneck.
安装
npx skills add NVIDIA/skills --skill jetson-speculative-decodingnpx skills add NVIDIA/skills --skill jetson-speculative-decoding -a claude-code -gnpx skills add NVIDIA/skills --skill jetson-speculative-decoding -a codex -ggemini skills install https://github.com/NVIDIA/skills.git --path skills/jetson-speculative-decodinggh skill install NVIDIA/skills jetson-speculative-decodinggit clone --depth 1 https://github.com/NVIDIA/skills /tmp/NVIDIA__skills
mkdir -p ~/.claude/skills ~/.agents/skills
cp -R /tmp/NVIDIA__skills/skills/jetson-speculative-decoding ~/.claude/skills/jetson-speculative-decoding # Claude Code
cp -R /tmp/NVIDIA__skills/skills/jetson-speculative-decoding ~/.agents/skills/jetson-speculative-decoding # Codex, Cursor, Gemini CLI, Copilot, OpenCode