trocr for Urdu
This model is a fine-tuned version of cxfajar197/urdu-ocr on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 2.2939
- Cer: 0.2622
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: 5e-05
- train_batch_size: 8
- eval_batch_size: 8
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 2
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Cer |
|---|---|---|---|---|
| 0.2374 | 0.1594 | 1000 | 2.4320 | 0.3063 |
| 0.2788 | 0.3188 | 2000 | 2.3795 | 0.3029 |
| 0.2845 | 0.4782 | 3000 | 2.3814 | 0.2694 |
| 0.2793 | 0.6377 | 4000 | 2.2703 | 0.2676 |
| 0.2735 | 0.7971 | 5000 | 2.2114 | 0.3016 |
| 0.2739 | 0.9565 | 6000 | 2.2326 | 0.3004 |
| 0.1781 | 1.1159 | 7000 | 2.2932 | 0.2810 |
| 0.1392 | 1.2753 | 8000 | 2.3545 | 0.2828 |
| 0.1252 | 1.4347 | 9000 | 2.3462 | 0.2515 |
| 0.1212 | 1.5941 | 10000 | 2.3429 | 0.2493 |
| 0.1172 | 1.7535 | 11000 | 2.2981 | 0.2769 |
| 0.1091 | 1.9130 | 12000 | 2.2939 | 0.2622 |
Framework versions
- Transformers 4.45.1
- Pytorch 2.4.0
- Datasets 3.0.1
- Tokenizers 0.20.0