vit-base_rvl_cdip-N1K_AURC_64
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: 0.3118
- Accuracy: 0.8952
- Brier Loss: 0.1766
- Nll: 0.8835
- F1 Micro: 0.8952
- F1 Macro: 0.8951
- Ece: 0.0747
- Aurc: 0.0206
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: 64
- eval_batch_size: 64
- 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 | 250 | 0.1770 | 0.8875 | 0.1709 | 1.2031 | 0.8875 | 0.8885 | 0.0519 | 0.0208 |
| 0.0228 | 2.0 | 500 | 0.2135 | 0.8852 | 0.1813 | 1.1542 | 0.8852 | 0.8853 | 0.0557 | 0.0228 |
| 0.0228 | 3.0 | 750 | 0.1750 | 0.8918 | 0.1729 | 1.0088 | 0.8918 | 0.8917 | 0.0628 | 0.0192 |
| 0.0066 | 4.0 | 1000 | 0.2117 | 0.8955 | 0.1697 | 0.9611 | 0.8955 | 0.8954 | 0.0655 | 0.0189 |
| 0.0066 | 5.0 | 1250 | 0.2578 | 0.8958 | 0.1714 | 0.9234 | 0.8958 | 0.8958 | 0.0690 | 0.0194 |
| 0.0021 | 6.0 | 1500 | 0.2752 | 0.8962 | 0.1730 | 0.9093 | 0.8962 | 0.8964 | 0.0709 | 0.0197 |
| 0.0021 | 7.0 | 1750 | 0.2949 | 0.8972 | 0.1748 | 0.8841 | 0.8972 | 0.8972 | 0.0708 | 0.0200 |
| 0.0014 | 8.0 | 2000 | 0.3037 | 0.8955 | 0.1755 | 0.8842 | 0.8955 | 0.8954 | 0.0739 | 0.0204 |
| 0.0014 | 9.0 | 2250 | 0.3045 | 0.8952 | 0.1764 | 0.8839 | 0.8952 | 0.8951 | 0.0741 | 0.0206 |
| 0.0013 | 10.0 | 2500 | 0.3118 | 0.8952 | 0.1766 | 0.8835 | 0.8952 | 0.8951 | 0.0747 | 0.0206 |
Framework versions
- Transformers 4.33.3
- Pytorch 2.2.0.dev20231002
- Datasets 2.7.1
- Tokenizers 0.13.3