{
  "model_name": "wav2vec2-audio-emotion-meld-seac-finetuned",
  "architecture": "Wav2Vec2 encoder (frozen) + Audio classification head",
  "task": "speech-emotion-recognition",
  "modality": "audio-only",
  "datasets": {
    "pretraining_dataset": "MELD",
    "finetuning_dataset": "SEAC"
  },
  "language": "Serbian",
  "labels": [
    "neutral",
    "joy",
    "anger",
    "sadness",
    "fear"
  ],
  "num_labels": 5,
  "encoder": {
    "type": "wav2vec2",
    "frozen": true,
    "pooling": "mean"
  },
  "training": {
    "loss": "weighted_cross_entropy",
    "optimizer": "AdamW",
    "early_stopping": true,
    "lr_scheduler": "ReduceLROnPlateau"
  },
  "evaluation": {
    "accuracy": 0.7107,
    "weighted_f1": 0.7130
  },
  "framework": "pytorch"
}