Whisper - Serbian Model
This model is a fine-tuned version of openai/whisper-medium on the common_voice_17_0 dataset. It achieves the following results on the evaluation set:
- Loss: 0.4195
- Wer: 20.5133
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 |
|---|---|---|---|---|
| 0.0368 | 4.6083 | 1000 | 0.2510 | 21.8831 |
| 0.0045 | 9.2166 | 2000 | 0.3441 | 20.3572 |
| 0.0002 | 13.8249 | 3000 | 0.4126 | 20.6520 |
| 0.0001 | 18.4332 | 4000 | 0.4195 | 20.5133 |
Framework versions
- Transformers 4.45.2
- Pytorch 2.5.1
- Datasets 3.1.0
- Tokenizers 0.20.1