trocr-bigram5-BY
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: 0.5955
- Cer: 0.0292
- Wer: 0.0838
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: 4650
- num_epochs: 3
Training results
| Training Loss | Epoch | Step | Validation Loss | Cer | Wer |
|---|---|---|---|---|---|
| 4.0314 | 0.1290 | 2000 | 1.7102 | 0.1973 | 0.3994 |
| 2.9555 | 0.2581 | 4000 | 1.4864 | 0.1415 | 0.3105 |
| 2.8902 | 0.3871 | 6000 | 1.2645 | 0.1040 | 0.2529 |
| 2.5335 | 0.5161 | 8000 | 1.1481 | 0.0841 | 0.2173 |
| 2.2147 | 0.6452 | 10000 | 1.0777 | 0.0772 | 0.2002 |
| 2.1987 | 0.7742 | 12000 | 0.9946 | 0.0666 | 0.1747 |
| 2.1294 | 0.9032 | 14000 | 0.9288 | 0.0605 | 0.1616 |
| 1.7200 | 1.0323 | 16000 | 0.8747 | 0.0541 | 0.1501 |
| 1.5009 | 1.1613 | 18000 | 0.8430 | 0.0500 | 0.1391 |
| 1.4939 | 1.2903 | 20000 | 0.7990 | 0.0451 | 0.1266 |
| 1.3924 | 1.4194 | 22000 | 0.7779 | 0.0426 | 0.1194 |
| 1.3807 | 1.5484 | 24000 | 0.7468 | 0.0431 | 0.1200 |
| 1.5015 | 1.6774 | 26000 | 0.7113 | 0.0355 | 0.1034 |
| 1.3585 | 1.8065 | 28000 | 0.6908 | 0.0350 | 0.1011 |
| 1.3697 | 1.9355 | 30000 | 0.6745 | 0.0359 | 0.1016 |
| 1.0135 | 2.0645 | 32000 | 0.6620 | 0.0346 | 0.0994 |
| 0.9966 | 2.1935 | 34000 | 0.6456 | 0.0349 | 0.0985 |
| 0.9714 | 2.3226 | 36000 | 0.6364 | 0.0324 | 0.0934 |
| 0.9632 | 2.4516 | 38000 | 0.6226 | 0.0307 | 0.0886 |
| 0.9263 | 2.5806 | 40000 | 0.6129 | 0.0303 | 0.0887 |
| 1.0649 | 2.7097 | 42000 | 0.6051 | 0.0298 | 0.0863 |
| 1.0234 | 2.8387 | 44000 | 0.5983 | 0.0285 | 0.0823 |
| 1.0024 | 2.9677 | 46000 | 0.5955 | 0.0293 | 0.0839 |
| 0.8827 | 3.0 | 46500 | 0.5955 | 0.0292 | 0.0838 |
Framework versions
- Transformers 5.15.0
- Pytorch 2.11.0+cu128
- Datasets 5.0.1
- Tokenizers 0.22.2