vit-small_rvl_cdip
This model is a fine-tuned version of WinKawaks/vit-small-patch16-224 on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.0854
- Accuracy: 0.9110
- Brier Loss: 0.1305
- Nll: 1.2339
- F1 Micro: 0.9110
- F1 Macro: 0.9112
- Ece: 0.0097
- Aurc: 0.0122
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: 0.0001
- train_batch_size: 128
- eval_batch_size: 128
- 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 |
|---|---|---|---|---|---|---|---|---|---|---|
| 0.2638 | 1.0 | 2500 | 0.2093 | 0.8574 | 0.2035 | 1.6580 | 0.8574 | 0.8564 | 0.0105 | 0.0272 |
| 0.1559 | 2.0 | 5000 | 0.1413 | 0.8854 | 0.1645 | 1.5062 | 0.8854 | 0.8863 | 0.0101 | 0.0185 |
| 0.1036 | 3.0 | 7500 | 0.1212 | 0.8994 | 0.1461 | 1.4353 | 0.8994 | 0.8999 | 0.0096 | 0.0150 |
| 0.0693 | 4.0 | 10000 | 0.1196 | 0.9013 | 0.1440 | 1.3827 | 0.9013 | 0.9017 | 0.0152 | 0.0148 |
| 0.0451 | 5.0 | 12500 | 0.1094 | 0.9062 | 0.1391 | 1.3099 | 0.9062 | 0.9064 | 0.0124 | 0.0139 |
| 0.0317 | 6.0 | 15000 | 0.0997 | 0.9073 | 0.1357 | 1.2889 | 0.9073 | 0.9078 | 0.0091 | 0.0132 |
| 0.0224 | 7.0 | 17500 | 0.0961 | 0.9081 | 0.1348 | 1.2705 | 0.9081 | 0.9082 | 0.0100 | 0.0129 |
| 0.0166 | 8.0 | 20000 | 0.0890 | 0.9099 | 0.1328 | 1.2484 | 0.9099 | 0.9102 | 0.0085 | 0.0126 |
| 0.0117 | 9.0 | 22500 | 0.0862 | 0.9096 | 0.1316 | 1.2428 | 0.9096 | 0.9100 | 0.0100 | 0.0123 |
| 0.0085 | 10.0 | 25000 | 0.0854 | 0.9110 | 0.1305 | 1.2339 | 0.9110 | 0.9112 | 0.0097 | 0.0122 |
Framework versions
- Transformers 4.26.1
- Pytorch 1.13.1.post200
- Datasets 2.9.0
- Tokenizers 0.13.2