bert-base-uncased-transformers-github-128
This model is a fine-tuned version of bert-base-uncased on the None dataset. It achieves the following results on the evaluation set:
- Loss: 1.2348
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: 16
- eval_batch_size: 8
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 16
Training results
| Training Loss | Epoch | Step | Validation Loss |
|---|---|---|---|
| 2.0247 | 1.0 | 582 | 1.6457 |
| 1.5989 | 2.0 | 1164 | 1.4157 |
| 1.4449 | 3.0 | 1746 | 1.3494 |
| 1.3579 | 4.0 | 2328 | 1.3774 |
| 1.3039 | 5.0 | 2910 | 1.1908 |
| 1.2428 | 6.0 | 3492 | 1.2780 |
| 1.19 | 7.0 | 4074 | 1.2569 |
| 1.1544 | 8.0 | 4656 | 1.1927 |
| 1.126 | 9.0 | 5238 | 1.1703 |
| 1.0893 | 10.0 | 5820 | 1.2100 |
| 1.0631 | 11.0 | 6402 | 1.1988 |
| 1.0417 | 12.0 | 6984 | 1.1643 |
| 1.0252 | 13.0 | 7566 | 1.2202 |
| 1.0101 | 14.0 | 8148 | 1.1678 |
| 0.9972 | 15.0 | 8730 | 1.0999 |
| 0.995 | 16.0 | 9312 | 1.2348 |
Framework versions
- Transformers 4.25.1
- Pytorch 1.13.0+cu116
- Datasets 2.8.0
- Tokenizers 0.13.2
- Downloads last month
- 3