lukealonso

/

MiniMax-M2.7-NVFP4

like 47

Using VLLM:Nightly cu130

CUDA_VISIBLE_DEVICES=0,1 \
SAFETENSORS_FAST_GPU=1 \
OMP_NUM_THREADS=8 \
NCCL_P2P_DISABLE=1 \
VLLM_USE_FLASHINFER_MOE_FP4=0 \
vllm serve lukealonso/MiniMax-M2.7-NVFP4 \
  --port 23333 \
  --trust-remote-code \
  --served-model-name MM-27 \
  --tensor-parallel-size 2 \
  --enable-auto-tool-choice \
  --tool-call-parser minimax_m2 \
  --reasoning-parser minimax_m2_append_think \
  --disable-custom-all-reduce \
  --enable_expert_parallel \
  --gpu-memory-utilization 0.95 \
  --max-num-seqs 8 \
  --kv-cache-dtype turboquant_k8v4 \
  --max-model-len 130k