Verified record · from the answer index
jetson-speculative-decoding
skilltested · static-verifiedscore 45/100signed passport ✓1,130 uses
Add EAGLE-3 or draft-model speculative decoding to a Jetson vLLM server when TPOT is the bottleneck.
Declared tools (1)
jetson-speculative-decoding
skill:nvidia/skills/jetson-speculative-decoding · published by nvidia · source repository