File size: 1,397 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": 5320,
"images": 1064,
"mean_seconds": 1.8277028220793219,
"median_seconds": 1.8259914284572005,
"std_seconds": 0.014033581550791104,
"ci95_seconds": 0.0003771106329976519,
"min_seconds": 1.7931781895458698,
"max_seconds": 1.9092448819428682,
"images_per_second": 0.5471348995688098,
"per_image_file": "/root/vepfs/Infinity/outputs/normal_eval/2026-06-17/12-38-50/sintel/ours/inference_times.json"
}
}
} |