bert_uncased_L-4_H-256_A-4_qqp
This model is a fine-tuned version of google/bert_uncased_L-4_H-256_A-4 on the GLUE QQP dataset. It achieves the following results on the evaluation set:
- Loss: 0.2840
- Accuracy: 0.8775
- F1: 0.8327
- Combined Score: 0.8551
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: 256
- eval_batch_size: 256
- seed: 10
- optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: linear
- num_epochs: 50
Training results
| Training Loss | Epoch | Step | Validation Loss | Accuracy | F1 | Combined Score |
|---|---|---|---|---|---|---|
| 0.3985 | 1.0 | 1422 | 0.3341 | 0.8486 | 0.7966 | 0.8226 |
| 0.3199 | 2.0 | 2844 | 0.3058 | 0.8636 | 0.8245 | 0.8440 |
| 0.2819 | 3.0 | 4266 | 0.2883 | 0.8732 | 0.8341 | 0.8536 |
| 0.2525 | 4.0 | 5688 | 0.2840 | 0.8775 | 0.8327 | 0.8551 |
| 0.2304 | 5.0 | 7110 | 0.2858 | 0.8808 | 0.8448 | 0.8628 |
| 0.2094 | 6.0 | 8532 | 0.2877 | 0.8817 | 0.8450 | 0.8633 |
| 0.1912 | 7.0 | 9954 | 0.2909 | 0.8823 | 0.8462 | 0.8642 |
| 0.1749 | 8.0 | 11376 | 0.2944 | 0.8856 | 0.8512 | 0.8684 |
| 0.1604 | 9.0 | 12798 | 0.3125 | 0.8863 | 0.8526 | 0.8694 |
Framework versions
- Transformers 4.46.3
- Pytorch 2.2.1+cu118
- Datasets 2.17.0
- Tokenizers 0.20.3