whisper-small-pashto-run10-augfromscratch
This model is a fine-tuned version of openai/whisper-small on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.6747
- Wer: 45.2848
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: 8
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 16
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 600
- training_steps: 21854
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 2.0828 | 0.0137 | 100 | 1.2287 | 69.7292 |
| 1.4691 | 0.0275 | 200 | 0.9251 | 58.5994 |
| 1.1827 | 0.0412 | 300 | 0.8029 | 56.3585 |
| 1.0427 | 0.0549 | 400 | 0.7679 | 53.5574 |
| 0.9808 | 0.0686 | 500 | 0.7501 | 52.9038 |
| 0.9841 | 0.0824 | 600 | 0.7365 | 52.7918 |
| 0.939 | 0.0961 | 700 | 0.7100 | 49.9907 |
| 0.9025 | 0.1098 | 800 | 0.7010 | 49.2997 |
| 0.8144 | 0.1235 | 900 | 0.6722 | 47.9739 |
| 0.7727 | 0.1373 | 1000 | 0.6786 | 47.6937 |
| 0.7464 | 0.1510 | 1100 | 0.6796 | 48.0859 |
| 0.7386 | 0.1647 | 1200 | 0.6566 | 47.0588 |
| 0.6685 | 0.1784 | 1300 | 0.6728 | 45.1727 |
| 0.6469 | 0.1922 | 1400 | 0.6608 | 48.9636 |
| 0.5535 | 0.2059 | 1500 | 0.6709 | 46.7974 |
| 0.597 | 0.2196 | 1600 | 0.6751 | 43.8282 |
| 0.5789 | 0.2334 | 1700 | 0.6673 | 44.5938 |
| 0.5856 | 0.2471 | 1800 | 0.6831 | 44.2204 |
| 0.5264 | 0.2608 | 1900 | 0.6661 | 44.3137 |
| 0.5485 | 0.2745 | 2000 | 0.6713 | 44.5191 |
| 0.4879 | 0.2883 | 2100 | 0.6747 | 45.2848 |
Framework versions
- Transformers 4.44.2
- Pytorch 2.4.0+cu118
- Datasets 2.19.0
- Tokenizers 0.19.1