whisper-small-finetuned-iwslt-irish-no-finetuning
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: 4.7121
- Wer: 99.6296
- Bleu: 2.9795
- Chrf++: 18.4960
- Comet: 0.4437
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 |
|---|---|---|---|---|---|---|---|
| 2.2805 | 2.6709 | 1250 | 4.3869 | 119.8862 | 2.1896 | 17.4135 | 0.4437 |
| 1.6995 | 5.3419 | 2500 | 4.4563 | 99.1236 | 3.0948 | 18.7809 | 0.4481 |
| 1.5895 | 8.0128 | 3750 | 4.6113 | 99.4489 | 2.7503 | 18.5821 | 0.4473 |
| 1.4975 | 10.6838 | 5000 | 4.7121 | 99.6296 | 2.9795 | 18.4960 | 0.4437 |
Framework versions
- Transformers 4.52.0
- Pytorch 2.12.1+cu130
- Datasets 2.21.0
- Tokenizers 0.21.4