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

62 lines
1.4 KiB
JSON

{
"candidate": {
"name": "rizzo_flow",
"model_id": "rizzo_flow@55633c8cbd2b",
"revision": "55633c8cbd2b826bd3eefdeb05310450996649df",
"adapter_version": "0.2.0",
"device": "cuda",
"context_limit": 2048,
"probabilities_source": "native"
},
"precision": "gguf-q8_0",
"device": "cuda",
"split": "test",
"examples": 1744,
"seed": 20261007,
"warmup": 5,
"batch_size": 1,
"state_budget_tokens": 512,
"process_guard": "ok",
"repetition": 3,
"load_seconds": 5.27,
"warmup_latency_ms": [
130.4,
101.0,
101.2,
76.8,
102.1
],
"latency_ms": {
"n": 1739,
"p50": 114.7081149999849,
"p95": 180.32949450000046,
"p99": 185.14998192000178,
"mean": 125.59644951063844
},
"errors": {},
"truncated": 0,
"with_diagnostics": 0,
"operational_env": {},
"state_tokens_max": 126,
"rss_peak_mib": 4532.5546875,
"gpu_mem_peak_mib": 5115.0,
"gpu_util_mean_pct": 74.09382151029749,
"quality": {
"n": 1744,
"errors": 0,
"accuracy": 0.8543577981651376,
"macro_f1": 0.8770666746856607,
"macro_precision": 0.858397796047086,
"macro_recall": 0.9122970471150125
},
"calibration_raw": {
"brier": 0.228059515858952,
"nll": 0.5092075237954721,
"ece": 0.03709107184248669
},
"oci": {
"region": "sa-saopaulo-1",
"run_id": "20261007T232719Z-rizzo_flow-evaluate"
}
}