bert_uncased_L-8_H-768_A-12_massive
This model is a fine-tuned version of google/bert_uncased_L-8_H-768_A-12 on the massive dataset. It achieves the following results on the evaluation set:
- Loss: 0.5735
- Accuracy: 0.8918
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: 64
- eval_batch_size: 64
- seed: 33
- distributed_type: multi-GPU
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 15
Training results
| Training Loss | Epoch | Step | Validation Loss | Accuracy |
|---|---|---|---|---|
| 2.0003 | 1.0 | 180 | 0.7856 | 0.8328 |
| 0.623 | 2.0 | 360 | 0.5288 | 0.8751 |
| 0.3368 | 3.0 | 540 | 0.4599 | 0.8844 |
| 0.2037 | 4.0 | 720 | 0.4858 | 0.8829 |
| 0.1257 | 5.0 | 900 | 0.5099 | 0.8839 |
| 0.0829 | 6.0 | 1080 | 0.5392 | 0.8854 |
| 0.0552 | 7.0 | 1260 | 0.5547 | 0.8893 |
| 0.0365 | 8.0 | 1440 | 0.5735 | 0.8918 |
| 0.0246 | 9.0 | 1620 | 0.5910 | 0.8854 |
| 0.0163 | 10.0 | 1800 | 0.5977 | 0.8883 |
| 0.0116 | 11.0 | 1980 | 0.6056 | 0.8874 |
| 0.0087 | 12.0 | 2160 | 0.6108 | 0.8874 |
| 0.0071 | 13.0 | 2340 | 0.6099 | 0.8903 |
| 0.0059 | 14.0 | 2520 | 0.6170 | 0.8903 |
| 0.0051 | 15.0 | 2700 | 0.6148 | 0.8908 |
Framework versions
- Transformers 4.34.0
- Pytorch 1.14.0a0+410ce96
- Datasets 2.14.5
- Tokenizers 0.14.1