mirror of
https://github.com/mcloh/decision-models-benchmark.git
synced 2026-11-05 22:04:12 +00:00
57 lines
1.2 KiB
JSON
57 lines
1.2 KiB
JSON
{
|
|
"candidate": {
|
|
"name": "laya",
|
|
"model_id": "laya@1720e3e3357c",
|
|
"revision": "1720e3e3357cfe1e281542e223f8273b0890ca34",
|
|
"adapter_version": "0.1.0",
|
|
"device": "cpu",
|
|
"context_limit": 1024,
|
|
"probabilities_source": "native"
|
|
},
|
|
"precision": "float32",
|
|
"device": "cpu",
|
|
"split": "dev",
|
|
"examples": 50,
|
|
"seed": 20261007,
|
|
"warmup": 5,
|
|
"batch_size": 1,
|
|
"state_budget_tokens": 512,
|
|
"process_guard": "ok",
|
|
"load_seconds": 9.29,
|
|
"warmup_latency_ms": [
|
|
240.3,
|
|
202.2,
|
|
115.4,
|
|
101.4,
|
|
164.7
|
|
],
|
|
"latency_ms": {
|
|
"n": 45,
|
|
"p50": 161.33673899997802,
|
|
"p95": 200.54640860000745,
|
|
"p99": 220.7323225200185,
|
|
"mean": 152.56732057777953
|
|
},
|
|
"errors": {},
|
|
"truncated": 0,
|
|
"state_tokens_max": 88,
|
|
"rss_peak_mib": 2673.00390625,
|
|
"quality": {
|
|
"n": 50,
|
|
"errors": 0,
|
|
"accuracy": 0.66,
|
|
"macro_f1": 0.6346938775510205,
|
|
"macro_precision": 0.6244047619047619,
|
|
"macro_recall": 0.7142857142857142
|
|
},
|
|
"calibration_raw": {
|
|
"brier": 0.5661517359520769,
|
|
"nll": 2.3105314537533026,
|
|
"ece": 0.2523094204409023
|
|
},
|
|
"oci": {
|
|
"region": "us-chicago-1",
|
|
"run_id": "20261007T174256Z-laya-evaluate"
|
|
}
|
|
}
|