risa-ai/sentiment-v2

Hyperparameters

        "epochs": 32,
        "learning_rate": 2e-5,
        "batch_size_train": 24,
        "batch_size_val": 16,
        "max_len": 512,
        "num_worker": 4,
        "max_grad_norm": 5,
    }

Run Summary Logs

   accuracy 0.8318462096481682
         f1 0.8314074021814951
  precision 0.7324855145009559
     recall 0.7052524442856924
 train_loss 0.006222551137038769

Evaluation Result

              precision    recall  f1-score   support

    confused       0.63      0.62      0.62      1170
    negative       0.73      0.60      0.66       515
      netral       0.89      0.90      0.89      9888
    positive       0.70      0.70      0.70      2212

    accuracy                           0.83     13785
   macro avg       0.73      0.70      0.72     13785
weighted avg       0.83      0.83      0.83     13785

Confusion Matrix