mirror of
https://github.com/mcloh/decision-models-benchmark.git
synced 2026-11-05 22:04:12 +00:00
59 lines
1.2 KiB
JSON
59 lines
1.2 KiB
JSON
{
|
|
"candidate": {
|
|
"name": "semif",
|
|
"model_id": "semif@851bf6e806ef",
|
|
"revision": "851bf6e806efd8d0a36b00ddf55e13ccb7b8cd0a",
|
|
"adapter_version": "0.1.0",
|
|
"device": "cuda",
|
|
"context_limit": 8192,
|
|
"probabilities_source": "native"
|
|
},
|
|
"precision": "bfloat16",
|
|
"device": "cuda",
|
|
"split": "dev",
|
|
"examples": 50,
|
|
"seed": 20261007,
|
|
"warmup": 5,
|
|
"batch_size": 1,
|
|
"state_budget_tokens": 512,
|
|
"process_guard": "ok",
|
|
"load_seconds": 6.7,
|
|
"warmup_latency_ms": [
|
|
432.5,
|
|
103.9,
|
|
75.2,
|
|
73.8,
|
|
106.7
|
|
],
|
|
"latency_ms": {
|
|
"n": 45,
|
|
"p50": 113.77530000001457,
|
|
"p95": 181.30440580000595,
|
|
"p99": 183.7473541199995,
|
|
"mean": 122.29987855555905
|
|
},
|
|
"errors": {},
|
|
"truncated": 0,
|
|
"state_tokens_max": 89,
|
|
"rss_peak_mib": 9019.65234375,
|
|
"gpu_mem_peak_mib": 8513.0,
|
|
"gpu_util_mean_pct": 45.68,
|
|
"quality": {
|
|
"n": 50,
|
|
"errors": 0,
|
|
"accuracy": 0.92,
|
|
"macro_f1": 0.9392857142857143,
|
|
"macro_precision": 0.9404761904761905,
|
|
"macro_recall": 0.9464285714285714
|
|
},
|
|
"calibration_raw": {
|
|
"brier": 0.1618760535683978,
|
|
"nll": 0.3303471264424595,
|
|
"ece": 0.1143176772925542
|
|
},
|
|
"oci": {
|
|
"region": "sa-saopaulo-1",
|
|
"run_id": "20261007T175428Z-semif-evaluate"
|
|
}
|
|
}
|