bart-kurd-spell-base
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: 0.1930
- Cer: 1.5424
- Wer: 8.3088
- Gen Len: 12.6945
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: 256
- eval_batch_size: 256
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 10.0
Training results
| Training Loss | Epoch | Step | Validation Loss | Cer | Wer | Gen Len |
|---|---|---|---|---|---|---|
| 0.4548 | 1.54 | 20000 | 0.4117 | 2.8856 | 13.6181 | 12.7807 |
| 0.2723 | 3.07 | 40000 | 0.2736 | 2.1004 | 10.5883 | 12.6808 |
| 0.2246 | 4.61 | 60000 | 0.2303 | 1.8035 | 9.4897 | 12.7048 |
| 0.1812 | 6.14 | 80000 | 0.2122 | 1.6804 | 8.9349 | 12.6937 |
| 0.1693 | 7.68 | 100000 | 0.2001 | 1.589 | 8.5464 | 12.7045 |
| 0.1498 | 9.22 | 120000 | 0.1942 | 1.5546 | 8.3598 | 12.6935 |
Framework versions
- Transformers 4.34.0.dev0
- Pytorch 1.13.1
- Datasets 2.14.5
- Tokenizers 0.13.3