vit-base_rvl_cdip-N1K_ce_256
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: 0.4495
- Accuracy: 0.8935
- Brier Loss: 0.1753
- Nll: 1.0235
- F1 Micro: 0.8935
- F1 Macro: 0.8937
- Ece: 0.0696
- Aurc: 0.0181
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: 256
- eval_batch_size: 256
- 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 | 63 | 0.3678 | 0.8972 | 0.1554 | 1.1865 | 0.8972 | 0.8975 | 0.0427 | 0.0165 |
| No log | 2.0 | 126 | 0.3774 | 0.896 | 0.1584 | 1.1527 | 0.8960 | 0.8962 | 0.0470 | 0.0170 |
| No log | 3.0 | 189 | 0.4050 | 0.892 | 0.1688 | 1.1092 | 0.892 | 0.8924 | 0.0578 | 0.0177 |
| No log | 4.0 | 252 | 0.4089 | 0.8945 | 0.1675 | 1.0874 | 0.8945 | 0.8948 | 0.0582 | 0.0177 |
| No log | 5.0 | 315 | 0.4255 | 0.8935 | 0.1704 | 1.0678 | 0.8935 | 0.8936 | 0.0640 | 0.0179 |
| No log | 6.0 | 378 | 0.4324 | 0.8945 | 0.1715 | 1.0540 | 0.8945 | 0.8948 | 0.0648 | 0.0179 |
| No log | 7.0 | 441 | 0.4404 | 0.894 | 0.1728 | 1.0302 | 0.894 | 0.8941 | 0.0672 | 0.0181 |
| 0.0579 | 8.0 | 504 | 0.4452 | 0.8932 | 0.1747 | 1.0316 | 0.8932 | 0.8934 | 0.0685 | 0.0180 |
| 0.0579 | 9.0 | 567 | 0.4479 | 0.8935 | 0.1749 | 1.0256 | 0.8935 | 0.8937 | 0.0693 | 0.0181 |
| 0.0579 | 10.0 | 630 | 0.4495 | 0.8935 | 0.1753 | 1.0235 | 0.8935 | 0.8937 | 0.0696 | 0.0181 |
Framework versions
- Transformers 4.33.3
- Pytorch 2.2.0.dev20231002
- Datasets 2.7.1
- Tokenizers 0.13.3