trocr-aug3track-250k
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: 0.3800
- Cer: 0.0230
- Wer: 0.0607
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 |
|---|---|---|---|---|---|
| 5.3301 | 0.1290 | 2000 | 2.4039 | 0.2468 | 0.4842 |
| 3.8025 | 0.2581 | 4000 | 1.8936 | 0.1654 | 0.3407 |
| 2.9084 | 0.3871 | 6000 | 1.4610 | 0.1164 | 0.2597 |
| 2.4395 | 0.5161 | 8000 | 1.1142 | 0.0888 | 0.2060 |
| 2.1606 | 0.6452 | 10000 | 1.0733 | 0.0800 | 0.1894 |
| 1.9200 | 0.7742 | 12000 | 0.8557 | 0.0608 | 0.1519 |
| 1.7398 | 0.9032 | 14000 | 0.7667 | 0.0563 | 0.1353 |
| 1.2903 | 1.0323 | 16000 | 0.6978 | 0.0498 | 0.1223 |
| 1.2324 | 1.1613 | 18000 | 0.6481 | 0.0450 | 0.1130 |
| 1.2523 | 1.2903 | 20000 | 0.6086 | 0.0406 | 0.1045 |
| 1.1177 | 1.4194 | 22000 | 0.5661 | 0.0381 | 0.0949 |
| 1.0789 | 1.5484 | 24000 | 0.5312 | 0.0358 | 0.0899 |
| 0.9692 | 1.6774 | 26000 | 0.5001 | 0.0316 | 0.0819 |
| 1.0177 | 1.8065 | 28000 | 0.4834 | 0.0316 | 0.0808 |
| 0.9540 | 1.9355 | 30000 | 0.4544 | 0.0277 | 0.0729 |
| 0.7212 | 2.0645 | 32000 | 0.4332 | 0.0274 | 0.0712 |
| 0.7048 | 2.1935 | 34000 | 0.4288 | 0.0272 | 0.0706 |
| 0.6208 | 2.3226 | 36000 | 0.4164 | 0.0270 | 0.0699 |
| 0.6726 | 2.4516 | 38000 | 0.4043 | 0.0254 | 0.0671 |
| 0.6400 | 2.5806 | 40000 | 0.3968 | 0.0239 | 0.0642 |
| 0.6058 | 2.7097 | 42000 | 0.3885 | 0.0241 | 0.0640 |
| 0.6313 | 2.8387 | 44000 | 0.3827 | 0.0233 | 0.0616 |
| 0.6330 | 2.9677 | 46000 | 0.3796 | 0.0229 | 0.0607 |
| 0.6001 | 3.0 | 46500 | 0.3800 | 0.0230 | 0.0607 |
Framework versions
- Transformers 5.15.0
- Pytorch 2.11.0+cu128
- Datasets 5.0.1
- Tokenizers 0.22.2