whisper-small-finetuned-iwslt-irish-welsh-curriculum-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: 3.4795
- Wer: 93.8291
- Bleu: 7.4317
- Chrf++: 27.4259
- Comet: 0.5182
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 |
|---|---|---|---|---|---|---|---|
| 2.2972 | 0.4264 | 2500 | 3.6835 | 140.0885 | 2.8386 | 20.0115 | 0.4674 |
| 2.0671 | 0.8528 | 5000 | 3.5530 | 127.0058 | 4.0401 | 22.7134 | 0.4854 |
| 2.2536 | 1.2792 | 7500 | 3.4554 | 110.1735 | 5.9475 | 25.9164 | 0.5103 |
| 2.0954 | 1.7056 | 10000 | 3.4357 | 99.0965 | 7.4245 | 26.9493 | 0.5168 |
| 2.0935 | 2.1320 | 12500 | 3.4540 | 97.4250 | 6.9609 | 26.8370 | 0.5199 |
| 2.0305 | 2.5584 | 15000 | 3.4795 | 93.8291 | 7.4317 | 27.4259 | 0.5182 |
Framework versions
- Transformers 4.52.0
- Pytorch 2.12.1+cu130
- Datasets 2.21.0
- Tokenizers 0.21.4