Whisper fine-tuned on FluencyBank — openai/whisper-large-v3-turbo
This model is a fine-tuned version of openai/whisper-large-v3-turbo on the FluencyBank Timestamped dataset. It achieves the following results on the evaluation set:
- Loss: 1.8510
- Wer: 9.9758
- Cer: 5.8628
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: 8e-06
- train_batch_size: 32
- eval_batch_size: 32
- seed: 42
- gradient_accumulation_steps: 4
- total_train_batch_size: 128
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_ratio: 0.1
- training_steps: 2500
- label_smoothing_factor: 0.1
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
|---|---|---|---|---|---|
| 1.4697 | 11.6279 | 250 | 1.7083 | 12.7005 | 6.4682 |
| 1.4366 | 23.2558 | 500 | 1.7474 | 9.8879 | 5.6989 |
| 1.4266 | 34.8837 | 750 | 1.7590 | 9.9978 | 5.8992 |
| 1.4248 | 46.5116 | 1000 | 1.7597 | 10.2395 | 6.0267 |
| 1.4195 | 58.1395 | 1250 | 1.8063 | 9.6902 | 5.6580 |
| 1.4183 | 69.7674 | 1500 | 1.8249 | 9.9978 | 5.8492 |
| 1.4176 | 81.3953 | 1750 | 1.8405 | 9.9319 | 5.8309 |
| 1.4173 | 93.0233 | 2000 | 1.8477 | 9.9978 | 5.8901 |
| 1.4172 | 104.6512 | 2250 | 1.8514 | 9.9539 | 5.8810 |
| 1.4172 | 116.2791 | 2500 | 1.8510 | 9.9758 | 5.8628 |
Framework versions
- Transformers 4.45.2
- Pytorch 2.10.0+cu128
- Datasets 4.0.0
- Tokenizers 0.20.3