{
  "model_name": "Qwen/Qwen3-4B",
  "gpu_name": "NVIDIA H100",
  "attention_implementation": "flash_attention_3_hopper",
  "num_requests": 10,
  "prompt_length": 512,
  "generation_length": 100,
  "prefill": {
    "stage_name": "prefill",
    "duration_ms": 323.99015384726226,
    "tokens_processed": 5120,
    "tokens_per_second": 15802.949377324925,
    "energy_joules": 17.092000007629395,
    "energy_per_token": 0.0033382812514901163,
    "avg_power_watts": 93.64442380045372,
    "peak_memory_gb": 46.02825927734375,
    "avg_gpu_util_percent": 40.0
  },
  "decode": {
    "stage_name": "decode",
    "duration_ms": 30513.75844143331,
    "tokens_processed": 1000,
    "tokens_per_second": 32.772101867403634,
    "energy_joules": 4915.5139999985695,
    "energy_per_token": 4.915513999998569,
    "avg_power_watts": 161.199160874206,
    "peak_memory_gb": 46.02825927734375,
    "avg_gpu_util_percent": 40.0
  },
  "e2e_latency_ms": 3083.7748595280573,
  "e2e_tokens_per_second": 198.4580677506596,
  "e2e_energy_joules": 4932.606000006199,
  "e2e_energy_per_token": 0.8059813725500325,
  "ttft_ms": 32.399015384726226,
  "itl_ms": 30.51375844143331,
  "timestamp": 1768541839.3186588
}