vit-base_rvl_cdip-N1K_ce_2
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: 1.1280
- Accuracy: 0.871
- Brier Loss: 0.2424
- Nll: 1.0979
- F1 Micro: 0.871
- F1 Macro: 0.8714
- Ece: 0.1202
- Aurc: 0.0321
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: 2
- eval_batch_size: 2
- 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 |
|---|---|---|---|---|---|---|---|---|---|---|
| 0.6617 | 1.0 | 8000 | 0.8187 | 0.8305 | 0.2858 | 1.3679 | 0.8305 | 0.8317 | 0.1182 | 0.0482 |
| 0.4446 | 2.0 | 16000 | 0.7450 | 0.8598 | 0.2427 | 1.2010 | 0.8598 | 0.8593 | 0.1100 | 0.0344 |
| 0.3853 | 3.0 | 24000 | 0.9422 | 0.8452 | 0.2771 | 1.3313 | 0.8452 | 0.8459 | 0.1288 | 0.0413 |
| 0.2259 | 4.0 | 32000 | 0.9828 | 0.8498 | 0.2702 | 1.3463 | 0.8498 | 0.8475 | 0.1267 | 0.0426 |
| 0.1824 | 5.0 | 40000 | 1.0204 | 0.8578 | 0.2603 | 1.2439 | 0.8578 | 0.8580 | 0.1263 | 0.0369 |
| 0.0343 | 6.0 | 48000 | 1.1306 | 0.8532 | 0.2716 | 1.1966 | 0.8532 | 0.8533 | 0.1340 | 0.0367 |
| 0.0288 | 7.0 | 56000 | 1.1081 | 0.8638 | 0.2552 | 1.2206 | 0.8638 | 0.8642 | 0.1243 | 0.0345 |
| 0.0122 | 8.0 | 64000 | 1.1533 | 0.8625 | 0.2579 | 1.2019 | 0.8625 | 0.8620 | 0.1260 | 0.0348 |
| 0.004 | 9.0 | 72000 | 1.1360 | 0.868 | 0.2487 | 1.0798 | 0.868 | 0.8691 | 0.1228 | 0.0330 |
| 0.0027 | 10.0 | 80000 | 1.1280 | 0.871 | 0.2424 | 1.0979 | 0.871 | 0.8714 | 0.1202 | 0.0321 |
Framework versions
- Transformers 4.33.3
- Pytorch 2.2.0.dev20231002
- Datasets 2.7.1
- Tokenizers 0.13.3