1t.yaml

- serve_id: vllm_model
  engine: vllm
  engine_args:
    model: /models/Kimi-K2-Instruct # add Post-trained models
    served_model_name: Kimi-K2-Instruct-nvidia-origin # add Post-trained models
    host: 0.0.0.0
    uvicorn_log_level: warning
    max_model_len: 32768
    port: 30000
  engine_args_specific:
    vllm:
      tensor_parallel_size: 16 # recommended by Moonshot https://modelscope.cn/models/moonshotai/Kimi-K2-Instruct/file/view/master/docs%2Fdeploy_guidance.md?status=1
      pipeline_parallel_size: 1
      gpu_memory_utilization: 0.98
      trust_remote_code: true
      enforce_eager: true # enforce_eager is recommended to be set true, false may trigger unknown but reproduced NCCL exception
      enable_chunked_prefill: true
  profile:
    prefix_len: 0
    input_len: 1024
    output_len: 1024
    num_prompts: 128
    range_ratio: 1