rocm_vllm_nightly/config.json

{
  "services": {
    "api": {
      "host": "0.0.0.0",
      "port": 8002
    },
    "openai": {
      "host": "0.0.0.0",
      "port": 8001
    }
  },
  "public_model_name": "Qwen_local_model",
  "default_enable_thinking": false,
  "reasoning_enabled": false,
  "api_key": "sk-szcjw",
  "tensor_parallel_size": 2,
  "dtype": "bfloat16",
  "model_root": "/opt/model",
  "offline_mode": true,
  "revision": "",
  "models": {
    "default": "Qwen3.5-35B-A3B-GPTQ-Int4",
    "selected": "Qwen3.6-27B-FP8",
    "profiles": {
      "Qwen3.6-35B-A3B-FP8": {
        "local_path": "Qwen3.6-35B-A3B-FP8",
        "dtype": "float16",
        "quantization": "fp8",
        "ctx": "262144",
        "max_tokens": "65536",
        "max_num_batched_tokens": 32768,
        "trust_remote": true,
        "enforce_eager": false,
        "valid_tp": [
          2
        ],
        "max_num_seqs": "256",
        "gpu_util": "0.92",
        "tool_call_parser": "qwen3_coder",
        "reasoning_parser": "qwen3",
        "enable_auto_tool_choice": true,
        "language_model_only": true,
        "served_model_name": "Qwen3.6-35B-A3B-FP8",
        "hf_model_id": "Qwen/Qwen3.6-35B-A3B-FP8"
      },
      "Qwen3.6-27B": {
        "local_path": "Qwen3.6-27B",
        "ctx": "131072",
        "max_tokens": "32768",
        "max_num_batched_tokens": 16384,
        "trust_remote": true,
        "enforce_eager": false,
        "valid_tp": [
          1,
          2
        ],
        "max_num_seqs": 32,
        "gpu_util": "0.96",
        "kv_cache_dtype": "fp8",
        "enable_prefix_caching": true,
        "tool_call_parser": "qwen3_coder",
        "reasoning_parser": "qwen3",
        "enable_auto_tool_choice": true,
        "language_model_only": true,
        "served_model_name": "Qwen3.6-27B",
        "hf_model_id": "Qwen/Qwen3.6-27B"
      },
      "Qwen3.6-27B-FP8": {
        "local_path": "Qwen3.6-27B-FP8",
        "dtype": "auto",
        "quantization": "fp8",
        "ctx": "131072",
        "max_tokens": "32768",
        "max_num_batched_tokens": 16384,
        "trust_remote": true,
        "enforce_eager": false,
        "valid_tp": [
          1,
          2
        ],
        "max_num_seqs": 32,
        "gpu_util": "0.90",
        "tool_call_parser": "qwen3_coder",
        "reasoning_parser": "qwen3",
        "enable_auto_tool_choice": true,
        "language_model_only": true,
        "served_model_name": "Qwen3.6-27B-FP8",
        "hf_model_id": "Qwen/Qwen3.6-27B-FP8",
        "env": {
          "VLLM_RPC_TIMEOUT": "300"
        },
        "speculative": {
          "method": "dflash",
          "model": "Qwen3.6-27B-DFlash",
          "num_speculative_tokens": 8,
          "draft_tensor_parallel_size": 2
        }
      }
    }
  }
}