whisper-small-finetuned-iwslt-irish-welsh-curriculum-b
This model is a fine-tuned version of openai/whisper-small on the None dataset. It achieves the following results on the evaluation set:
- Loss: 4.7344
- Wer: 104.2103
- Bleu: 3.1336
- Chrf++: 18.6533
- Comet: 0.4427
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: 16
- eval_batch_size: 8
- seed: 42
- optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 1000
- training_steps: 5000
- mixed_precision_training: Native AMP
- label_smoothing_factor: 0.1
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer | Bleu | Chrf++ | Comet |
|---|---|---|---|---|---|---|---|
| 2.4756 | 2.6709 | 1250 | 4.3221 | 112.7394 | 2.3960 | 17.9712 | 0.4415 |
| 1.6684 | 5.3419 | 2500 | 4.4973 | 105.9902 | 2.8872 | 18.8111 | 0.4461 |
| 1.5494 | 8.0128 | 3750 | 4.6734 | 104.0567 | 2.9919 | 18.6236 | 0.4477 |
| 1.485 | 10.6838 | 5000 | 4.7344 | 104.2103 | 3.1336 | 18.6533 | 0.4427 |
Framework versions
- Transformers 4.52.0
- Pytorch 2.12.1+cu130
- Datasets 2.21.0
- Tokenizers 0.21.4