{
  "benchmarked_devices": [
    "cuda"
  ],
  "parameters": {
    "total_params": 20200546,
    "trainable_params": 18378970,
    "total_params_m": 20.200546,
    "trainable_params_m": 18.37897
  },
  "cuda": {
    "memory": {
      "baseline_memory_mb": 314.5068359375,
      "peak_memory_mb": 321.7763671875,
      "current_memory_mb": 315.08154296875,
      "inference_memory_mb": 7.26953125,
      "device": "cuda"
    },
    "inference_single": {
      "batch_size": 1,
      "mean_latency_ms": 11.141735980272642,
      "std_latency_ms": 0.04346183998720415,
      "median_latency_ms": 11.13254650044837,
      "p95_latency_ms": 11.162551251254627,
      "p99_latency_ms": 11.36283751031442,
      "min_latency_ms": 11.105517995019909,
      "max_latency_ms": 11.442682000051718,
      "latency_per_image_ms": 11.141735980272642,
      "throughput_img_per_sec": 89.7526203969096,
      "device": "cuda"
    },
    "inference_batch32": {
      "batch_size": 32,
      "mean_latency_ms": 41.95240021988866,
      "std_latency_ms": 0.12925345881619055,
      "median_latency_ms": 41.9302680020337,
      "p95_latency_ms": 41.98958034867246,
      "p99_latency_ms": 42.48272620359785,
      "min_latency_ms": 41.89117100031581,
      "max_latency_ms": 42.82661800243659,
      "latency_per_image_ms": 1.3110125068715206,
      "throughput_img_per_sec": 762.7692297049917,
      "device": "cuda"
    }
  }
}