whisper-large-v2-greek
This model is a fine-tuned version of openai/whisper-large-v2 on the fleurs dataset. It achieves the following results on the evaluation set:
- Loss: 0.2734
- Wer Ortho: 0.2102
- Wer: 0.1774
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: 2e-05
- train_batch_size: 2
- eval_batch_size: 2
- seed: 42
- gradient_accumulation_steps: 16
- total_train_batch_size: 32
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: constant_with_warmup
- lr_scheduler_warmup_steps: 50
- num_epochs: 7
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer Ortho | Wer |
|---|---|---|---|---|---|
| 0.1809 | 1.0 | 274 | 0.2244 | 0.2261 | 0.1947 |
| 0.0977 | 2.0 | 549 | 0.2306 | 0.2204 | 0.1856 |
| 0.0594 | 3.0 | 824 | 0.2332 | 0.2137 | 0.1814 |
| 0.0454 | 4.0 | 1099 | 0.2667 | 0.2315 | 0.1985 |
| 0.028 | 5.0 | 1374 | 0.2579 | 0.2151 | 0.1822 |
| 0.022 | 6.0 | 1649 | 0.2674 | 0.2188 | 0.1863 |
| 0.0202 | 6.98 | 1918 | 0.2734 | 0.2102 | 0.1774 |
Framework versions
- Transformers 4.30.0.dev0
- Pytorch 2.0.1+cu117
- Datasets 2.13.1
- Tokenizers 0.13.3