start_sglang.sh

export HSA_FORCE_FINE_GRAIN_PCIE=1
export HIP_VISIBLE_DEVICES=0,1,2,3,4,5,6,7
export NCCL_IB_HCA=mlx5_0:1
export SGLANG_ALLOW_OVERWRITE_LONGER_CONTEXT_LEN=1

python3 -m sglang.launch_server \
    --model Qwen/Qwen3-Next-80B-A3B-Instruct \
    --tp 4 \
    --dtype bfloat16 \
    --port 8001