vit-base_rvl_cdip-N1K_aAURC_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: 0.5297
- Accuracy: 0.874
- Brier Loss: 0.2289
- Nll: 0.9943
- F1 Micro: 0.874
- F1 Macro: 0.8744
- Ece: 0.1117
- Aurc: 0.0291
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 |
|---|---|---|---|---|---|---|---|---|---|---|
| 0.2467 | 1.0 | 4000 | 0.3863 | 0.8403 | 0.2519 | 1.2558 | 0.8403 | 0.8407 | 0.0998 | 0.0394 |
| 0.1931 | 2.0 | 8000 | 0.4295 | 0.8482 | 0.2575 | 1.2140 | 0.8482 | 0.8486 | 0.1120 | 0.0362 |
| 0.1278 | 3.0 | 12000 | 0.4308 | 0.86 | 0.2406 | 1.1212 | 0.8600 | 0.8601 | 0.1063 | 0.0332 |
| 0.0798 | 4.0 | 16000 | 0.5079 | 0.853 | 0.2588 | 1.2528 | 0.853 | 0.8523 | 0.1221 | 0.0348 |
| 0.0422 | 5.0 | 20000 | 0.5064 | 0.8638 | 0.2443 | 1.1013 | 0.8638 | 0.8635 | 0.1165 | 0.0315 |
| 0.0123 | 6.0 | 24000 | 0.5186 | 0.8672 | 0.2378 | 1.0551 | 0.8672 | 0.8668 | 0.1155 | 0.0328 |
| 0.0048 | 7.0 | 28000 | 0.5372 | 0.8752 | 0.2306 | 1.1080 | 0.8752 | 0.8756 | 0.1101 | 0.0310 |
| 0.0098 | 8.0 | 32000 | 0.5395 | 0.8732 | 0.2325 | 1.0344 | 0.8732 | 0.8732 | 0.1135 | 0.0306 |
| 0.0019 | 9.0 | 36000 | 0.5249 | 0.875 | 0.2283 | 1.0203 | 0.875 | 0.8751 | 0.1099 | 0.0290 |
| 0.002 | 10.0 | 40000 | 0.5297 | 0.874 | 0.2289 | 0.9943 | 0.874 | 0.8744 | 0.1117 | 0.0291 |
Framework versions
- Transformers 4.33.3
- Pytorch 2.2.0.dev20231002
- Datasets 2.7.1
- Tokenizers 0.13.3