distilbert-base-uncased-finetuned-yelp-reviews
This model is a fine-tuned version of distilbert-base-uncased on the yelp_review_full dataset. It achieves the following results on the evaluation set:
- Loss: 0.8288
- Accuracy: 0.6418
- F1: 0.6425
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: 2e-05
- train_batch_size: 32
- eval_batch_size: 32
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 2
Training results
| Training Loss | Epoch | Step | Validation Loss | Accuracy | F1 |
|---|---|---|---|---|---|
| 0.8991 | 1.0 | 1524 | 0.8396 | 0.6302 | 0.6294 |
| 0.754 | 2.0 | 3048 | 0.8288 | 0.6418 | 0.6425 |
Framework versions
- Transformers 4.21.2
- Pytorch 1.12.1+cu113
- Datasets 2.4.0
- Tokenizers 0.12.1