vit-base_rvl_cdip-N1K_aAURC_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: 0.5215
- Accuracy: 0.888
- Brier Loss: 0.1918
- Nll: 0.9026
- F1 Micro: 0.888
- F1 Macro: 0.8883
- Ece: 0.0880
- Aurc: 0.0205
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 |
|---|---|---|---|---|---|---|---|---|---|---|
| 0.1629 | 1.0 | 500 | 0.3779 | 0.8875 | 0.1721 | 1.1899 | 0.8875 | 0.8877 | 0.0531 | 0.0201 |
| 0.1234 | 2.0 | 1000 | 0.4074 | 0.8868 | 0.1790 | 1.1333 | 0.8868 | 0.8874 | 0.0647 | 0.0213 |
| 0.0616 | 3.0 | 1500 | 0.4257 | 0.888 | 0.1813 | 1.0677 | 0.888 | 0.8879 | 0.0695 | 0.0201 |
| 0.0303 | 4.0 | 2000 | 0.4595 | 0.885 | 0.1869 | 1.0256 | 0.885 | 0.8856 | 0.0776 | 0.0222 |
| 0.0133 | 5.0 | 2500 | 0.4902 | 0.8848 | 0.1922 | 0.9983 | 0.8848 | 0.8849 | 0.0831 | 0.0228 |
| 0.0083 | 6.0 | 3000 | 0.4941 | 0.8862 | 0.1903 | 0.9464 | 0.8862 | 0.8868 | 0.0850 | 0.0211 |
| 0.0051 | 7.0 | 3500 | 0.5116 | 0.8875 | 0.1928 | 0.9118 | 0.8875 | 0.8873 | 0.0875 | 0.0207 |
| 0.0043 | 8.0 | 4000 | 0.5154 | 0.8882 | 0.1910 | 0.9138 | 0.8882 | 0.8887 | 0.0864 | 0.0205 |
| 0.0041 | 9.0 | 4500 | 0.5221 | 0.8865 | 0.1924 | 0.9101 | 0.8865 | 0.8868 | 0.0896 | 0.0206 |
| 0.0037 | 10.0 | 5000 | 0.5215 | 0.888 | 0.1918 | 0.9026 | 0.888 | 0.8883 | 0.0880 | 0.0205 |
Framework versions
- Transformers 4.33.3
- Pytorch 2.2.0.dev20231002
- Datasets 2.7.1
- Tokenizers 0.13.3