whisper-small-finetuned-iwslt-irish-synthetic
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: 3.4815
- Wer: 94.9404
- Bleu: 9.1198
- Chrf++: 29.5535
- Comet: 0.5382
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: 15000
- mixed_precision_training: Native AMP
- label_smoothing_factor: 0.1
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer | Bleu | Chrf++ | Comet |
|---|---|---|---|---|---|---|---|
| 3.2981 | 0.4264 | 2500 | 3.5015 | 109.1615 | 5.0736 | 23.5316 | 0.4935 |
| 2.5927 | 0.8528 | 5000 | 3.4141 | 118.1334 | 5.9677 | 26.7723 | 0.5084 |
| 2.3202 | 1.2792 | 7500 | 3.3988 | 100.2530 | 7.7452 | 28.1924 | 0.5260 |
| 2.1779 | 1.7056 | 10000 | 3.4073 | 100.1897 | 8.2058 | 28.8603 | 0.5372 |
| 2.0624 | 2.1320 | 12500 | 3.4574 | 90.2873 | 9.3718 | 29.4620 | 0.5372 |
| 1.922 | 2.5584 | 15000 | 3.4815 | 94.9404 | 9.1198 | 29.5535 | 0.5382 |
Framework versions
- Transformers 4.52.0
- Pytorch 2.12.1+cu130
- Datasets 2.21.0
- Tokenizers 0.21.4