vit-base_rvl_cdip-N1K_AURC_4
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: 0.2768
- Accuracy: 0.8738
- Brier Loss: 0.2167
- Nll: 0.9821
- F1 Micro: 0.8738
- F1 Macro: 0.8749
- Ece: 0.0970
- Aurc: 0.0292
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: 4
- eval_batch_size: 4
- 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 |
|---|---|---|---|---|---|---|---|---|---|---|
| 0.1764 | 1.0 | 4000 | 0.3808 | 0.8217 | 0.2750 | 1.2675 | 0.8217 | 0.8194 | 0.1016 | 0.0461 |
| 0.1131 | 2.0 | 8000 | 0.3321 | 0.8413 | 0.2583 | 1.3120 | 0.8413 | 0.8421 | 0.0949 | 0.0418 |
| 0.113 | 3.0 | 12000 | 0.3781 | 0.8207 | 0.2910 | 1.4889 | 0.8207 | 0.8213 | 0.1162 | 0.0496 |
| 0.0814 | 4.0 | 16000 | 0.4793 | 0.8157 | 0.3036 | 1.4208 | 0.8157 | 0.8151 | 0.1302 | 0.0552 |
| 0.0542 | 5.0 | 20000 | 0.2914 | 0.8658 | 0.2279 | 1.1541 | 0.8658 | 0.8657 | 0.0955 | 0.0320 |
| 0.0238 | 6.0 | 24000 | 0.3059 | 0.8568 | 0.2401 | 1.1686 | 0.8568 | 0.8581 | 0.1012 | 0.0354 |
| 0.0197 | 7.0 | 28000 | 0.3077 | 0.8545 | 0.2390 | 1.1659 | 0.8545 | 0.8553 | 0.1059 | 0.0354 |
| 0.0116 | 8.0 | 32000 | 0.3169 | 0.8705 | 0.2172 | 1.0323 | 0.8705 | 0.8704 | 0.0918 | 0.0314 |
| 0.0054 | 9.0 | 36000 | 0.2850 | 0.8738 | 0.2199 | 1.0171 | 0.8738 | 0.8747 | 0.0960 | 0.0302 |
| 0.0128 | 10.0 | 40000 | 0.2768 | 0.8738 | 0.2167 | 0.9821 | 0.8738 | 0.8749 | 0.0970 | 0.0292 |
Framework versions
- Transformers 4.33.3
- Pytorch 2.2.0.dev20231002
- Datasets 2.7.1
- Tokenizers 0.13.3