Files
decision-models-benchmark/reports/test/summary-gliner-cuda-r3.json
2026-10-07 20:53:56 -03:00

62 lines
1.4 KiB
JSON

{
"candidate": {
"name": "gliner",
"model_id": "gliner@a35a0cd3b7a0",
"revision": "a35a0cd3b7a0f00f2effc576f454cd48fa98aa5f",
"adapter_version": "0.2.0",
"device": "cuda",
"context_limit": 4096,
"probabilities_source": "native"
},
"precision": "float32",
"device": "cuda",
"split": "test",
"examples": 1744,
"seed": 20261007,
"warmup": 5,
"batch_size": 1,
"state_budget_tokens": 512,
"process_guard": "ok",
"repetition": 3,
"load_seconds": 21.84,
"warmup_latency_ms": [
240.4,
113.1,
106.3,
274.5,
23.1
],
"latency_ms": {
"n": 1739,
"p50": 21.212821000005988,
"p95": 31.739651800023694,
"p99": 33.962657579994584,
"mean": 21.807412646349036
},
"errors": {},
"truncated": 0,
"with_diagnostics": 0,
"operational_env": {},
"state_tokens_max": 136,
"rss_peak_mib": 3254.2734375,
"gpu_mem_peak_mib": 1807.0,
"gpu_util_mean_pct": 45.84251968503937,
"quality": {
"n": 1744,
"errors": 0,
"accuracy": 0.569954128440367,
"macro_f1": 0.5794749896128905,
"macro_precision": 0.7287383692866102,
"macro_recall": 0.562777593678045
},
"calibration_raw": {
"brier": 0.5973161312700161,
"nll": 1.3014290752544617,
"ece": 0.09860073873976281
},
"oci": {
"region": "sa-saopaulo-1",
"run_id": "20261007T234215Z-gliner-evaluate"
}
}