{
  "model_name": "Qwen/Qwen3-4B",
  "gpu_name": "NVIDIA H200",
  "attention_implementation": "flash_attention_3_hopper",
  "num_requests": 10,
  "prompt_length": 512,
  "generation_length": 100,
  "prefill": {
    "stage_name": "prefill",
    "duration_ms": 323.8773119999223,
    "tokens_processed": 5120,
    "tokens_per_second": 15808.455270868828,
    "energy_joules": 98.1449999999968,
    "energy_per_token": 0.019168945312499373,
    "avg_power_watts": 250.96736239598317,
    "peak_memory_gb": 46.1302490234375,
    "avg_gpu_util_percent": 32.2
  },
  "decode": {
    "stage_name": "decode",
    "duration_ms": 30558.618001000013,
    "tokens_processed": 1000,
    "tokens_per_second": 32.72399294913388,
    "energy_joules": 4828.459999999999,
    "energy_per_token": 4.828459999999999,
    "avg_power_watts": 157.61927190444868,
    "peak_memory_gb": 46.1302490234375,
    "avg_gpu_util_percent": 32.2
  },
  "e2e_latency_ms": 3088.2495312999936,
  "e2e_tokens_per_second": 198.17051497855476,
  "e2e_energy_joules": 4926.604999999996,
  "e2e_energy_per_token": 0.8050008169934634,
  "ttft_ms": 32.38773119999223,
  "itl_ms": 30.558618001000013,
  "timestamp": 1768541964.4743361
}