{
  "config": {
    "audio_model": "facebook/wav2vec2-base",
    "text_model": "roberta-base",
    "fusion": "cross_attention",
    "learning_rate": "1e-05",
    "batch_size": "8",
    "num_epochs": "7",
    "max_audio_length": "80000",
    "max_text_length": "128",
    "fusion_dim": "256",
    "dropout": "0.3",
    "freeze_encoders": "False",
    "output_dir": "./multimodal-emotion-model",
    "hub_model_id": "hugoaslm/multimodal-emotion-recognition",
    "push_to_hub": "True",
    "gradient_accumulation_steps": "4",
    "warmup_ratio": "0.1",
    "weight_decay": "0.01",
    "seed": "42",
    "mode": "multimodal",
    "run_name": "final-cross-attn-ls01",
    "label_smoothing": "0.1"
  },
  "val_metrics": {
    "eval_loss": 0.42387160658836365,
    "eval_accuracy": 0.8144072036018009,
    "eval_f1_weighted": 0.814478521287098,
    "eval_f1_macro": 0.7991105813540713,
    "eval_f1_angry": 0.8958677685950414,
    "eval_f1_calm": 0.59375,
    "eval_f1_disgust": 0.8183306055646481,
    "eval_f1_fear": 0.7622377622377622,
    "eval_f1_happy": 0.8248772504091653,
    "eval_f1_neutral": 0.8297213622291022,
    "eval_f1_sad": 0.7382753403933434,
    "eval_f1_surprise": 0.9298245614035088,
    "eval_runtime": 11.0916,
    "eval_samples_per_second": 180.226,
    "eval_steps_per_second": 22.54,
    "epoch": 7.0
  },
  "test_metrics": {
    "accuracy": 0.852760736196319,
    "f1_weighted": 0.8518738111646259,
    "f1_macro": 0.8508017687676436,
    "f1_angry": 0.8666666666666667,
    "f1_calm": 0.9230769230769231,
    "f1_disgust": 0.8163265306122449,
    "f1_fear": 0.819672131147541,
    "f1_happy": 0.8727272727272727,
    "f1_neutral": 0.8571428571428571,
    "f1_sad": 0.8
  },
  "classification_report": "              precision    recall  f1-score   support\n\n       angry     0.8667    0.8667    0.8667        30\n     disgust     0.9474    0.9000    0.9231        20\n        fear     0.9091    0.7407    0.8163        27\n       happy     0.8929    0.7576    0.8197        33\n     neutral     0.8000    0.9600    0.8727        25\n         sad     0.7826    0.9474    0.8571        19\n    surprise     0.7273    0.8889    0.8000         9\n\n    accuracy                         0.8528       163\n   macro avg     0.8466    0.8659    0.8508       163\nweighted avg     0.8612    0.8528    0.8519       163\n"
}