File size: 1,394 Bytes
4d35876 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 | {
"protocol": "Batch-size-1 per-image model inference latency measured inside each predictor with time.perf_counter and CUDA synchronization around the model-call section only. Each image uses unmeasured warmup iterations followed by measured repeats. Model loading, process startup, dataset export, image file I/O, metric computation, visualization writing, and result serialization are excluded.",
"methods": {
"ours": {
"enabled": true,
"available": true,
"method": "ours",
"protocol": "Batch-size-1 per-image model inference latency measured inside each predictor with time.perf_counter and CUDA synchronization around the model-call section only. Each image uses unmeasured warmup iterations followed by measured repeats. Model loading, process startup, dataset export, image file I/O, metric computation, visualization writing, and result serialization are excluded.",
"timed_observations": 500,
"images": 100,
"mean_seconds": 1.8394116687588393,
"median_seconds": 1.8351657567545772,
"std_seconds": 0.017275431719267167,
"ci95_seconds": 0.0015142583548655479,
"min_seconds": 1.8134799133986235,
"max_seconds": 1.8864054400473833,
"images_per_second": 0.5436520910377608,
"per_image_file": "/root/vepfs/Infinity/outputs/normal_eval/2026-06-17/12-17-47/ibims/ours/inference_times.json"
}
}
} |