#! /bin/bash

export CUDA_VISIBLE_DEVICES="4,5"
#export PYTHONWARNINGS="ignore::UserWarning:vllm.model_executor.layers.fla"
MODEL="/home/kongcunliang/workspace/pretrained-models/Qwen/Qwen3.5-27B-FP8"

vllm serve $MODEL \
  --served-model-name "Qwen3.5-27B" \
  --port 8000 \
  --tensor-parallel-size 2 \
  --max-model-len 16384 \
  --reasoning-parser qwen3 \
  --speculative-config '{"method":"mtp","num_speculative_tokens":1}'
# --enable-prefix-caching \