trocr-bigram4-BY-150k
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: 0.8047
- Cer: 0.0456
- Wer: 0.1252
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: 2775
- num_epochs: 3
Training results
| Training Loss | Epoch | Step | Validation Loss | Cer | Wer |
|---|---|---|---|---|---|
| 3.8458 | 0.2162 | 2000 | 1.6327 | 0.1750 | 0.3731 |
| 3.3301 | 0.4324 | 4000 | 1.4336 | 0.1252 | 0.2916 |
| 2.6102 | 0.6486 | 6000 | 1.2517 | 0.0961 | 0.2384 |
| 2.4748 | 0.8649 | 8000 | 1.1428 | 0.0794 | 0.2094 |
| 1.9151 | 1.0811 | 10000 | 1.0864 | 0.0780 | 0.1955 |
| 1.7786 | 1.2973 | 12000 | 1.0002 | 0.0685 | 0.1794 |
| 1.6665 | 1.5135 | 14000 | 0.9675 | 0.0574 | 0.1572 |
| 1.6515 | 1.7297 | 16000 | 0.9092 | 0.0578 | 0.1524 |
| 1.6083 | 1.9459 | 18000 | 0.8814 | 0.0513 | 0.1419 |
| 1.2360 | 2.1622 | 20000 | 0.8622 | 0.0492 | 0.1351 |
| 1.2638 | 2.3784 | 22000 | 0.8422 | 0.0509 | 0.1372 |
| 1.2311 | 2.5946 | 24000 | 0.8205 | 0.0488 | 0.1319 |
| 1.1231 | 2.8108 | 26000 | 0.8094 | 0.0465 | 0.1273 |
| 1.2555 | 3.0 | 27750 | 0.8047 | 0.0456 | 0.1252 |
Framework versions
- Transformers 5.15.0
- Pytorch 2.11.0+cu128
- Datasets 5.0.1
- Tokenizers 0.22.2