{
  "dataset": "GoEmotions",
  "source": "https://github.com/google-research/google-research/tree/master/goemotions",
  "task": "emotion",
  "n_labels": 27,
  "label_wording": "27 GoEmotions labels (neutral-only rows skipped; frozen names)",
  "train_rows": 30587,
  "calib_rows": 2000,
  "test_rows": 3821,
  "protocol": "Multi-label: every example sees all 27 emotion candidates. Official test untouched. Dev (capped) used only for temperature calibration.",
  "onesystem_version": "0.3.0",
  "device": "mps",
  "threshold": 0.3,
  "current_checkpoint": {
    "n": 3821,
    "n_labels": 27,
    "threshold": 0.3,
    "subset_accuracy": 0.0,
    "micro_f1": 0.0844,
    "macro_f1": 0.0827,
    "latency": {
      "n": 50,
      "p50_ms": 9.35,
      "p95_ms": 11.78,
      "mean_ms": 9.69
    },
    "seconds": 13.2
  },
  "finetune_split": {
    "train_rows": 30587,
    "calib_rows": 2000,
    "calib_source": "official_dev"
  },
  "domain_adapted": {
    "n": 3821,
    "n_labels": 27,
    "threshold": 0.3,
    "subset_accuracy": 0.4672,
    "micro_f1": 0.6482,
    "macro_f1": 0.6095,
    "latency": {
      "n": 50,
      "p50_ms": 11.77,
      "p95_ms": 17.89,
      "mean_ms": 11.91
    },
    "seconds": 14.1,
    "temperature": 1.0,
    "calib_nll": null,
    "train_run": {
      "steps": 7648,
      "history": [
        {
          "epoch": 1,
          "loss": 0.1921
        },
        {
          "epoch": 2,
          "loss": 0.0675
        }
      ],
      "seconds": 3521.7
    }
  }
}