metrics.json

{
  "temperature": 0.1278,
  "test_accuracy_raw": 0.9386,
  "test_macro_f1_raw": 0.9376,
  "test_macro_f1_with_reject_as_unknown": 0.9513,
  "coverage": 0.9206,
  "ece_10_bin": 0.0205,
  "multiclass_brier": 0.0827,
  "baseline_test_macro_f1": 0.8282
}