bart-kurd-spell-base-sn
This model is a fine-tuned version of facebook/bart-base on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.4815
- Cer: 2.1669
- Wer: 12.1294
- Bleu: 78.2542
- Chrf: 95.7354
- Gen Len: 16.7779
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: 5e-05
- train_batch_size: 64
- eval_batch_size: 64
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 5.0
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Cer | Wer | Bleu | Chrf | Gen Len |
|---|---|---|---|---|---|---|---|---|
| 0.2972 | 1.0 | 177897 | 0.5591 | 2.836 | 14.9407 | 73.3024 | 94.1403 | 16.7054 |
| 0.233 | 2.0 | 355794 | 0.5157 | 2.4613 | 13.4362 | 75.8819 | 95.0077 | 16.7604 |
| 0.2043 | 3.0 | 533691 | 0.4918 | 2.307 | 12.7609 | 77.0962 | 95.3849 | 16.7681 |
| 0.1753 | 4.0 | 711588 | 0.4871 | 2.2105 | 12.3386 | 77.928 | 95.6297 | 16.7765 |
| 0.1655 | 5.0 | 889485 | 0.4815 | 2.1669 | 12.1294 | 78.2542 | 95.7354 | 16.7779 |
Framework versions
- Transformers 4.36.0.dev0
- Pytorch 2.1.0
- Datasets 2.15.0
- Tokenizers 0.15.0