File size: 1,386 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": 140,
"images": 28,
"mean_seconds": 1.828554026569639,
"median_seconds": 1.8316096141934395,
"std_seconds": 0.0386702388094181,
"ci95_seconds": 0.006405734104784912,
"min_seconds": 1.765766154974699,
"max_seconds": 1.9089088086038828,
"images_per_second": 0.5468802045056315,
"per_image_file": "/root/vepfs/Infinity/outputs/normal_eval/2026-06-17/12-17-47/toy/ours/inference_times.json"
}
}
} |