trocr-small-1style-hh104-stage1-frozen
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: 4.6194
- Cer: 0.7898
- Wer: 1.0518
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: 16
- eval_batch_size: 16
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 32
- 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: 0.1
- num_epochs: 2
Training results
| Training Loss | Epoch | Step | Validation Loss | Cer | Wer |
|---|---|---|---|---|---|
| 21.0359 | 0.2581 | 2000 | 5.1897 | 0.7888 | 1.0248 |
| 20.3343 | 0.5161 | 4000 | 4.9947 | 0.8164 | 1.1070 |
| 19.6582 | 0.7742 | 6000 | 4.8657 | 0.8185 | 1.1002 |
| 18.2319 | 1.0323 | 8000 | 4.7899 | 0.7948 | 1.0401 |
| 18.0070 | 1.2903 | 10000 | 4.7354 | 0.7906 | 1.0587 |
| 17.9656 | 1.5484 | 12000 | 4.6782 | 0.7821 | 1.0497 |
| 17.6443 | 1.8065 | 14000 | 4.6363 | 0.7858 | 1.0472 |
| 17.7350 | 2.0 | 15500 | 4.6194 | 0.7898 | 1.0518 |
Framework versions
- Transformers 5.14.1
- Pytorch 2.10.0+cu128
- Datasets 5.0.0
- Tokenizers 0.22.2