trocr-small-BY5-bridgeinit-selfinit
This model is a fine-tuned version of cyttic/trocr-hebrew-small-untrained on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 6.0410
- Cer: 0.7729
- Wer: 1.1879
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
- gradient_accumulation_steps: 2
- total_train_batch_size: 16
- optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 1550
- num_epochs: 1.0
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Cer | Wer |
|---|---|---|---|---|---|
| 13.5321 | 0.1290 | 2000 | 6.5331 | 0.7890 | 1.2575 |
| 12.6771 | 0.2581 | 4000 | 6.3124 | 0.7675 | 1.1423 |
| 12.8559 | 0.3871 | 6000 | 6.2843 | 0.7982 | 1.2820 |
| 12.5482 | 0.5161 | 8000 | 6.1586 | 0.7657 | 1.1255 |
| 12.3941 | 0.6452 | 10000 | 6.1102 | 0.7802 | 1.1589 |
| 12.5326 | 0.7742 | 12000 | 6.0781 | 0.7771 | 1.1789 |
| 12.4375 | 0.9032 | 14000 | 6.0520 | 0.7742 | 1.2098 |
| 12.1724 | 1.0 | 15500 | 6.0410 | 0.7729 | 1.1879 |
Framework versions
- Transformers 5.15.0
- Pytorch 2.10.0+cu128
- Datasets 5.0.1
- Tokenizers 0.22.2