vit-base_rvl_cdip-N1K_AURC_32
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: 0.3439
- Accuracy: 0.8962
- Brier Loss: 0.1805
- Nll: 0.8184
- F1 Micro: 0.8962
- F1 Macro: 0.8963
- Ece: 0.0767
- Aurc: 0.0220
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: 32
- eval_batch_size: 32
- 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 |
|---|---|---|---|---|---|---|---|---|---|---|
| 0.0301 | 1.0 | 500 | 0.1897 | 0.8808 | 0.1804 | 1.1636 | 0.8808 | 0.8807 | 0.0528 | 0.0227 |
| 0.0229 | 2.0 | 1000 | 0.2504 | 0.883 | 0.1834 | 1.1357 | 0.883 | 0.8832 | 0.0573 | 0.0248 |
| 0.0081 | 3.0 | 1500 | 0.2251 | 0.8858 | 0.1787 | 1.0242 | 0.8858 | 0.8858 | 0.0653 | 0.0221 |
| 0.004 | 4.0 | 2000 | 0.3075 | 0.886 | 0.1831 | 0.9279 | 0.886 | 0.8850 | 0.0744 | 0.0227 |
| 0.0023 | 5.0 | 2500 | 0.2491 | 0.8908 | 0.1791 | 0.9302 | 0.8907 | 0.8916 | 0.0728 | 0.0212 |
| 0.0014 | 6.0 | 3000 | 0.3067 | 0.8925 | 0.1795 | 0.8631 | 0.8925 | 0.8929 | 0.0752 | 0.0215 |
| 0.0012 | 7.0 | 3500 | 0.3277 | 0.8925 | 0.1812 | 0.8729 | 0.8925 | 0.8922 | 0.0764 | 0.0218 |
| 0.0009 | 8.0 | 4000 | 0.3386 | 0.895 | 0.1797 | 0.8406 | 0.895 | 0.8951 | 0.0760 | 0.0219 |
| 0.0007 | 9.0 | 4500 | 0.3383 | 0.8968 | 0.1808 | 0.8293 | 0.8968 | 0.8969 | 0.0747 | 0.0220 |
| 0.0006 | 10.0 | 5000 | 0.3439 | 0.8962 | 0.1805 | 0.8184 | 0.8962 | 0.8963 | 0.0767 | 0.0220 |
Framework versions
- Transformers 4.33.3
- Pytorch 2.2.0.dev20231002
- Datasets 2.7.1
- Tokenizers 0.13.3