roberta-sentiment
RoBERTa est un modèle d'analyse sentimentale développé par Facebook AI. Il est basé
sur l'architecture des transformers et est pré-entraîné sur une grande quantité de
données variées. RoBERTa est capable de comprendre et prédire avec précision le ton
émotionnel (positif, négatif ou neutre) d'un texte.
Model description
More information needed
Intended uses & limitations
More information needed
Training and evaluation data
More information needed
Training procedure
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 1e-05
- train_batch_size: 64
- eval_batch_size: 64
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 50
- num_epochs: 10
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss |
|---|
| 0.5451 | 1.0 | 713 | 0.5422 |
| 0.4785 | 2.0 | 1426 | 0.5585 |
| 0.4199 | 3.0 | 2139 | 0.5785 |
| 0.3608 | 4.0 | 2852 | 0.6038 |
| 0.3117 | 5.0 | 3565 | 0.6713 |
| 0.2684 | 6.0 | 4278 | 0.7366 |
| 0.2403 | 7.0 | 4991 | 0.7737 |
| 0.2137 | 8.0 | 5704 | 0.8276 |
| 0.1926 | 9.0 | 6417 | 0.8597 |
| 0.1778 | 10.0 | 7130 | 0.8863 |
Framework versions
- Transformers 4.25.1
- Pytorch 1.13.0+cu116
- Datasets 2.8.0
- Tokenizers 0.13.2
