<!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. -->
Whisper Large V2
This model is a fine-tuned version of openai/whisper-large-v2 on the Jasmin-CGN dataset (Group 5: native adults above 65). It achieves the following results on the evaluation set:
- Loss: 0.3321
- Wer: 13.0920
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: 3e-05
- train_batch_size: 16
- eval_batch_size: 8
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 25
- training_steps: 250
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 0.6708 | 0.47 | 25 | 0.3483 | 38.0428 |
| 0.3161 | 0.94 | 50 | 0.2836 | 15.0162 |
| 0.1862 | 1.42 | 75 | 0.2742 | 12.3425 |
| 0.1531 | 1.89 | 100 | 0.2689 | 13.4163 |
| 0.0895 | 2.36 | 125 | 0.2845 | 12.8475 |
| 0.066 | 2.83 | 150 | 0.2876 | 13.4056 |
| 0.0432 | 3.3 | 175 | 0.3130 | 12.9485 |
| 0.0257 | 3.77 | 200 | 0.3066 | 12.2203 |
| 0.015 | 4.25 | 225 | 0.3195 | 12.6987 |
| 0.0102 | 4.72 | 250 | 0.3321 | 13.0920 |
Framework versions
- Transformers 4.35.0.dev0
- Pytorch 2.1.0+cu121
- Datasets 2.14.5
- Tokenizers 0.14.1
