Whisper Small new-ru-pl AIIA
This model is a fine-tuned version of Maks545curve/whisper-small-new-ru-a on the Common Voice 17 dataset. It achieves the following results on the evaluation set:
- Loss: 0.3382
- Wer Ortho: 25.7888
- Wer: 16.3796
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: 3e-05
- train_batch_size: 32
- eval_batch_size: 32
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: constant_with_warmup
- lr_scheduler_warmup_steps: 70
- training_steps: 1000
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer Ortho | Wer |
|---|---|---|---|---|---|
| 0.344 | 0.5336 | 500 | 0.3786 | 29.1204 | 18.2225 |
| 0.2073 | 1.0672 | 1000 | 0.3382 | 25.7888 | 16.3796 |
Framework versions
- Transformers 4.41.2
- Pytorch 2.3.1+cpu
- Datasets 2.20.0
- Tokenizers 0.19.1