Update vLLM deployment section
#2
by rogerwyf - opened
vLLM now reads mamba ssm dtype from model's config for Qwen3.5 (https://huggingface.co/Qwen/Qwen3.5-397B-A17B/blob/main/config.json#L101), therefore passing dtype override is no longer necessary.
Also made a suggestion to simplify language model only mode
jklj077 changed pull request status to closed