Serve language and vision models on Jetson.
What it does
Provides Jetson-specific commands to serve a language or vision-language model, with an endpoint and a verification step.
When to use it
Use when asked to host a model on Jetson or start a vLLM server for an application.
How to use it
Give it a model to serve on Jetson; it returns a complete launch recipe and performs live checks or deployment only when requested.
What you provide
- a model identifier or checkpoint path
Uses
NJ
Good to know
- cannot be undone: vllm
- acts without asking
Latest review
What you need · 6
Live deployment requires a Jetson host, accessed directly or through a shell with access to its GPU runtime.
The container deployment path requires Docker with the NVIDIA GPU runtime exposed.
Access · 4
This skill
vllm
Execute
Starts a vLLM model server
Write
Removes an existing vllm container (irreversible)
Jetson
Write
Changes Jetson power and clocks