whisper-small-0DB
This model is a fine-tuned version of openai/whisper-small on the None dataset. It achieves the following results on the evaluation set:
- Cer: 34.8762
- Loss: 0.5518
- Wer: 77.7023
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: 1e-05
- train_batch_size: 32
- eval_batch_size: 16
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 64
- optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED 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: 100
- num_epochs: 10
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Cer | Validation Loss | Wer |
|---|---|---|---|---|---|
| 2.9880 | 1.2675 | 100 | 47.1725 | 0.7525 | 93.3292 |
| 1.3953 | 2.5350 | 200 | 37.4019 | 0.4301 | 83.2613 |
| 0.9869 | 3.8025 | 300 | 34.8762 | 0.3891 | 80.0494 |
| 0.5308 | 5.0637 | 400 | 34.7857 | 0.4148 | 78.7523 |
| 0.3279 | 6.3312 | 500 | 33.9505 | 0.4685 | 78.1347 |
| 0.1994 | 7.5987 | 600 | 34.8360 | 0.5106 | 78.6905 |
| 0.1139 | 8.8662 | 700 | 34.8762 | 0.5518 | 77.7023 |
Framework versions
- Transformers 5.0.0
- Pytorch 2.10.0+cu128
- Datasets 4.8.3
- Tokenizers 0.22.2