whisper-medium-reverse-ml-mft-1-exp
This model is a fine-tuned version of openai/whisper-medium on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.0743
- Wer: 39.9559
- Cer: 11.9281
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: 0.0002
- train_batch_size: 64
- eval_batch_size: 32
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 128
- optimizer: Use adamw_torch_fused with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: warmup_stable_decay
- lr_scheduler_warmup_steps: 593
- training_steps: 5928
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
|---|---|---|---|---|---|
| 0.0553 | 1.0 | 1482 | 0.0791 | 45.0545 | 14.0901 |
| 0.0476 | 2.0 | 2964 | 0.0705 | 42.7779 | 12.9730 |
| 0.0414 | 3.0 | 4446 | 0.0715 | 41.7625 | 12.6873 |
| 0.0265 | 4.0 | 5928 | 0.0743 | 39.9559 | 11.9281 |
Framework versions
- Transformers 4.51.1
- Pytorch 2.8.0+cu128
- Datasets 4.4.1
- Tokenizers 0.21.4