trocr-webfonts3-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: 0.5777
- Cer: 0.0293
- Wer: 0.0844
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 |
|---|---|---|---|---|---|
| 3.2769 | 0.1290 | 2000 | 1.4291 | 0.1362 | 0.3151 |
| 3.0838 | 0.2581 | 4000 | 1.2826 | 0.1023 | 0.2589 |
| 2.6336 | 0.3871 | 6000 | 1.2115 | 0.0891 | 0.2299 |
| 2.3281 | 0.5161 | 8000 | 1.0646 | 0.0745 | 0.1939 |
| 2.1610 | 0.6452 | 10000 | 0.9843 | 0.0656 | 0.1744 |
| 1.8384 | 0.7742 | 12000 | 0.9269 | 0.0563 | 0.1573 |
| 1.8396 | 0.9032 | 14000 | 0.8723 | 0.0532 | 0.1472 |
| 1.4733 | 1.0323 | 16000 | 0.8289 | 0.0524 | 0.1421 |
| 1.4156 | 1.1613 | 18000 | 0.7869 | 0.0457 | 0.1308 |
| 1.3125 | 1.2903 | 20000 | 0.7629 | 0.0397 | 0.1158 |
| 1.3227 | 1.4194 | 22000 | 0.7421 | 0.0463 | 0.1272 |
| 1.1755 | 1.5484 | 24000 | 0.7152 | 0.0388 | 0.1120 |
| 1.2515 | 1.6774 | 26000 | 0.6867 | 0.0364 | 0.1056 |
| 1.2560 | 1.8065 | 28000 | 0.6627 | 0.0337 | 0.1001 |
| 1.1380 | 1.9355 | 30000 | 0.6397 | 0.0301 | 0.0910 |
| 0.9784 | 2.0645 | 32000 | 0.6300 | 0.0327 | 0.0939 |
| 0.9256 | 2.1935 | 34000 | 0.6214 | 0.0328 | 0.0930 |
| 0.8967 | 2.3226 | 36000 | 0.6117 | 0.0323 | 0.0917 |
| 0.9117 | 2.4516 | 38000 | 0.6033 | 0.0311 | 0.0895 |
| 0.8049 | 2.5806 | 40000 | 0.5968 | 0.0309 | 0.0877 |
| 0.9319 | 2.7097 | 42000 | 0.5890 | 0.0288 | 0.0840 |
| 0.9467 | 2.8387 | 44000 | 0.5799 | 0.0296 | 0.0852 |
| 0.9058 | 2.9677 | 46000 | 0.5776 | 0.0295 | 0.0849 |
| 0.8585 | 3.0 | 46500 | 0.5777 | 0.0293 | 0.0844 |
Framework versions
- Transformers 5.15.0
- Pytorch 2.11.0+cu128
- Datasets 5.0.1
- Tokenizers 0.22.2