super_large_finetune_CM01
This model was trained from scratch on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 7.2285
- Wer: 0.7714
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: 0.0001
- train_batch_size: 15
- eval_batch_size: 8
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 857
- num_epochs: 50
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 1.0031 | 5.0 | 1715 | 1.9766 | 0.7857 |
| 0.2107 | 10.0 | 3430 | 3.8748 | 0.8238 |
| 0.1393 | 15.0 | 5145 | 4.7403 | 0.7952 |
| 0.0931 | 20.0 | 6860 | 3.5077 | 0.6667 |
| 0.0649 | 25.0 | 8575 | 7.7419 | 0.9333 |
| 0.0592 | 30.0 | 10290 | 5.6440 | 0.7762 |
| 0.0396 | 35.0 | 12005 | 6.9629 | 0.6810 |
| 0.03 | 40.0 | 13720 | 7.8282 | 0.7524 |
| 0.0191 | 45.0 | 15435 | 6.4626 | 0.7429 |
| 0.0121 | 50.0 | 17150 | 7.2285 | 0.7714 |
Framework versions
- Transformers 4.18.0
- Pytorch 1.10.2+cu102
- Datasets 2.3.2
- Tokenizers 0.12.1