vit-base_rvl_cdip-N1K_ce_128
This model is a fine-tuned version of jordyvl/vit-base_rvl-cdip on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.4776
- Accuracy: 0.8912
- Brier Loss: 0.1798
- Nll: 0.9844
- F1 Micro: 0.8912
- F1 Macro: 0.8915
- Ece: 0.0768
- Aurc: 0.0189
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: 128
- eval_batch_size: 128
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_ratio: 0.1
- num_epochs: 10
Training results
| Training Loss | Epoch | Step | Validation Loss | Accuracy | Brier Loss | Nll | F1 Micro | F1 Macro | Ece | Aurc |
|---|---|---|---|---|---|---|---|---|---|---|
| No log | 1.0 | 125 | 0.3896 | 0.893 | 0.1649 | 1.1887 | 0.893 | 0.8933 | 0.0484 | 0.0175 |
| No log | 2.0 | 250 | 0.3908 | 0.8948 | 0.1606 | 1.1433 | 0.8948 | 0.8950 | 0.0499 | 0.0179 |
| No log | 3.0 | 375 | 0.4188 | 0.892 | 0.1708 | 1.0860 | 0.892 | 0.8923 | 0.0607 | 0.0184 |
| 0.0953 | 4.0 | 500 | 0.4268 | 0.892 | 0.1707 | 1.0788 | 0.892 | 0.8924 | 0.0654 | 0.0184 |
| 0.0953 | 5.0 | 625 | 0.4414 | 0.8938 | 0.1719 | 1.0502 | 0.8938 | 0.8941 | 0.0664 | 0.0187 |
| 0.0953 | 6.0 | 750 | 0.4570 | 0.8932 | 0.1754 | 1.0253 | 0.8932 | 0.8936 | 0.0714 | 0.0187 |
| 0.0953 | 7.0 | 875 | 0.4681 | 0.891 | 0.1779 | 1.0018 | 0.891 | 0.8912 | 0.0752 | 0.0191 |
| 0.0128 | 8.0 | 1000 | 0.4720 | 0.8902 | 0.1792 | 0.9789 | 0.8902 | 0.8905 | 0.0771 | 0.0188 |
| 0.0128 | 9.0 | 1125 | 0.4757 | 0.8918 | 0.1794 | 0.9865 | 0.8918 | 0.8920 | 0.0760 | 0.0188 |
| 0.0128 | 10.0 | 1250 | 0.4776 | 0.8912 | 0.1798 | 0.9844 | 0.8912 | 0.8915 | 0.0768 | 0.0189 |
Framework versions
- Transformers 4.33.3
- Pytorch 2.2.0.dev20231002
- Datasets 2.7.1
- Tokenizers 0.13.3