Table 16. Best subset of hyperparameters for each model in multi-label sentiment classification.
Model | Epoch | Batch size | Weight decay | Learning rate |
---|---|---|---|---|
BETO | 10 | 8 | 0.195014 | 0.000028 |
ALBETO | 8 | 8 | 0.039639 | 0.000028 |
DistilBETO | 9 | 16 | 0.073405 | 0.000038 |
MarIA | 10 | 8 | 0.195014 | 0.000028 |
BERTIN | 10 | 8 | 0.081163 | 0.000017 |