metadata
tags:
- generated_from_trainer
datasets:
- tweet_eval
model-index:
- name: roberta-sentiment-analysis-finetune
results: []
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