{
  "version": 1,
  "profiles": [
    {
      "id": "qwen-gpu",
      "alias": "multi-gpu-smoke",
      "model": "qwen3.8-27b-ud-q4_k_xl",
      "runtime": "llama_cuda",
      "executable": "/home/philipbodenbach/.local/share/werk1112/backends/llama-cuda/upstream-ec928150501c/build/bin/llama-server",
      "build": "ec928150501c",
      "gpus": [
        "GPU-b41bf955-b7e6-6f64-82b2-6b3cf08450e9"
      ],
      "strategy": "single",
      "context": 2048,
      "batch": 256,
      "parallel": 1,
      "cpu_threads": 8,
      "memory": [
        {
          "weights": 18500000000,
          "kv": 268435456,
          "compute": 536870912,
          "overhead": 268435456,
          "reserve": 536870912
        }
      ],
      "host_bytes": 24000000000,
      "native_args": [
        "--no-warmup"
      ]
    }
  ]
}