← All Claude SkillsClaude Skill · Development
Jetson Speculative Decoding (vLLM)
Add EAGLE-3 or draft-model speculative decoding to a Jetson vLLM server when TPOT is the bottleneck.
Source: github.com/NVIDIA/skills/tree/HEAD/skills/jetson-speculative-decoding · License: Other
The full guide, versions and download are available to members — membership is free.