trocr-fonts1
This model is a fine-tuned version of cyttic/exp2-frozen-benyehuda-cont on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.4665
- Cer: 0.0246
- Wer: 0.0689
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: 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: 0.1
- num_epochs: 2
Training results
| Training Loss | Epoch | Step | Validation Loss | Cer | Wer |
|---|---|---|---|---|---|
| 3.8193 | 0.1290 | 2000 | 1.8610 | 0.1587 | 0.3426 |
| 2.9061 | 0.2581 | 4000 | 1.2697 | 0.0950 | 0.2268 |
| 2.3763 | 0.3871 | 6000 | 1.0212 | 0.0722 | 0.1773 |
| 1.9058 | 0.5161 | 8000 | 0.8708 | 0.0618 | 0.1508 |
| 1.8305 | 0.6452 | 10000 | 0.7672 | 0.0512 | 0.1316 |
| 1.5210 | 0.7742 | 12000 | 0.6915 | 0.0433 | 0.1134 |
| 1.4128 | 0.9032 | 14000 | 0.6434 | 0.0389 | 0.1039 |
| 1.0751 | 1.0323 | 16000 | 0.5982 | 0.0361 | 0.0972 |
| 1.0369 | 1.1613 | 18000 | 0.5833 | 0.0330 | 0.0892 |
| 1.0608 | 1.2903 | 20000 | 0.5430 | 0.0299 | 0.0834 |
| 0.9312 | 1.4194 | 22000 | 0.5184 | 0.0300 | 0.0798 |
| 1.0019 | 1.5484 | 24000 | 0.5002 | 0.0276 | 0.0752 |
| 0.8871 | 1.6774 | 26000 | 0.4833 | 0.0263 | 0.0733 |
| 0.8420 | 1.8065 | 28000 | 0.4747 | 0.0251 | 0.0708 |
| 0.8935 | 1.9355 | 30000 | 0.4670 | 0.0252 | 0.0694 |
| 0.7684 | 2.0 | 31000 | 0.4665 | 0.0246 | 0.0689 |
Framework versions
- Transformers 5.9.0
- Pytorch 2.11.0+cu128
- Datasets 5.0.1
- Tokenizers 0.22.2