PromptDustcache low
vLLM stores compilation artifacts and derived model data here; it holds no conversation content (requests are served in memory).
Where it lives
- all
~/.cache/vllmdir
What you can do
- Clearing this cache does not affect stored models; downloaded weights live under ~/.cache/huggingface/hub.
Source: docs.vllm.ai
This describes where vLLM stores data. Metadata only, never your content. Confidence: likely. Spot a mistake? Tell us.