Bert_Classifier
This model is a fine-tuned version of bert-base-cased on the yelp_review_full dataset. It achieves the following results on the evaluation set:
- Loss: 1.1067
- Accuracy: 0.5533
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: 5e-05
- train_batch_size: 8
- eval_batch_size: 8
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 3
Training results
| Training Loss | Epoch | Step | Validation Loss | Accuracy |
|---|---|---|---|---|
| No log | 1.0 | 188 | 1.0636 | 0.5 |
| No log | 2.0 | 376 | 1.0405 | 0.52 |
| 0.9962 | 3.0 | 564 | 1.1067 | 0.5533 |
Framework versions
- Transformers 4.21.3
- Pytorch 1.12.1+cu113
- Datasets 2.4.0
- Tokenizers 0.12.1