trocr-bigram2-ep3
This model is a fine-tuned version of cyttic/trocr-noise-bigram2 on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.4678
- Cer: 0.0269
- Wer: 0.0731
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: 1
Training results
| Training Loss | Epoch | Step | Validation Loss | Cer | Wer |
|---|---|---|---|---|---|
| 1.0919 | 0.1290 | 2000 | 0.6478 | 0.0386 | 0.1045 |
| 1.1127 | 0.2581 | 4000 | 0.6151 | 0.0378 | 0.0975 |
| 1.0402 | 0.3871 | 6000 | 0.5755 | 0.0334 | 0.0907 |
| 0.9637 | 0.5161 | 8000 | 0.5434 | 0.0327 | 0.0867 |
| 0.9027 | 0.6452 | 10000 | 0.5157 | 0.0318 | 0.0834 |
| 0.8803 | 0.7742 | 12000 | 0.4868 | 0.0284 | 0.0781 |
| 0.8560 | 0.9032 | 14000 | 0.4738 | 0.0278 | 0.0751 |
| 0.8372 | 1.0 | 15500 | 0.4678 | 0.0269 | 0.0731 |
Framework versions
- Transformers 5.9.0
- Pytorch 2.11.0+cu128
- Datasets 5.0.1
- Tokenizers 0.22.2