{
  "calibration": {
    "compute_probe_shape": [
      2048,
      2048,
      2048
    ],
    "effective_compute_tflops": 11.687860430783342,
    "effective_memory_gbps": 180.7487936816152,
    "memory_probe_mib": 64
  },
  "contrast_workload": "gemm_tiny",
  "device": {
    "backend": "pytorch",
    "device": "cuda:0",
    "dtype": "fp16",
    "name": "NVIDIA GeForce RTX 3050 Laptop GPU"
  },
  "impact": {
    "reason": "A measured regression result is required.",
    "status": "not_evaluated"
  },
  "measurement": {
    "measured_iterations": 50,
    "warmup_iterations": 10
  },
  "primary_workload": "gemm_large_square",
  "regression": {
    "reason": "No defensible performance-changing historical baseline was available.",
    "status": "not_evaluated"
  },
  "right_sizing": {
    "reason": "Only one physical device was measured.",
    "status": "omitted"
  },
  "runtime": {
    "cuda_runtime_version": "13.0",
    "pytorch_version": "2.14.1+cu130"
  },
  "schema_version": 2,
  "telemetry": {
    "sample_count": 35,
    "status": "captured",
    "supported_fields": [
      "clock_event_reasons",
      "gpu_utilization_percent",
      "memory_activity_percent",
      "memory_clock_mhz",
      "memory_used_bytes",
      "performance_state",
      "power_watts",
      "sm_clock_mhz",
      "temperature_celsius"
    ],
    "unavailable_fields": [
      "dram_bandwidth_utilization_percent",
      "sm_activity_percent",
      "sm_occupancy_percent",
      "tensor_activity_percent"
    ]
  },
  "tensorforge_commit": "37871229d6bc44991a4900ad00449fe390065b88",
  "validation": {
    "ape_percent": 5.8557024132649325,
    "measured_p50_seconds": 0.00019516450265655294,
    "measured_p95_seconds": 0.00022512414871016516,
    "predicted_bottleneck": "compute-bound",
    "predicted_seconds": 0.00018373625016465667
  }
}