Whisper Small Ru ORD 0.4 - Mizoru
This model is a fine-tuned version of openai/whisper-small on the ORD_0.4 dataset. It achieves the following results on the evaluation set:
- Loss: 1.0157
- Wer: 47.4481
- Cer: 27.9353
- Clean Wer: 38.8041
- Clean Cer: 22.5075
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: 1e-05
- train_batch_size: 16
- 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: 500
- training_steps: 4000
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer | Cer | Clean Wer | Clean Cer |
|---|---|---|---|---|---|---|---|
| 1.0606 | 1.0 | 573 | 1.0674 | 50.9989 | 29.8675 | 41.9998 | 24.3296 |
| 0.9142 | 2.0 | 1146 | 1.0090 | 48.2571 | 28.1474 | 39.9670 | 22.7534 |
| 0.7999 | 3.0 | 1719 | 1.0020 | 48.5513 | 28.3688 | 39.3516 | 22.9446 |
| 0.6655 | 4.0 | 2292 | 1.0157 | 47.4481 | 27.9353 | 38.8041 | 22.5075 |
| 0.5702 | 5.0 | 2865 | 1.0444 | 48.1203 | 28.1053 | 39.3807 | 22.5587 |
| 0.4817 | 6.0 | 3438 | 1.0735 | 47.8894 | 27.9447 | 39.2281 | 22.3609 |
Framework versions
- Transformers 4.39.3
- Pytorch 2.1.2
- Datasets 2.17.0
- Tokenizers 0.15.2
- Downloads last month
- 6