120min_whisper_small_FT_
This model is a fine-tuned version of openai/whisper-small on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.9793
- Wer: 73.6089
- Cer: 35.0299
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: 5e-05
- train_batch_size: 2
- eval_batch_size: 4
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 4
- 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: 400
- training_steps: 1500
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
|---|---|---|---|---|---|
| 1.3278 | 0.3817 | 100 | 1.0068 | 70.3920 | 26.3073 |
| 1.011 | 0.7634 | 200 | 0.9279 | 69.9889 | 28.5998 |
| 0.783 | 1.1450 | 300 | 0.9639 | 70.0364 | 28.7620 |
| 0.6481 | 1.5267 | 400 | 1.0095 | 76.7942 | 36.6462 |
| 0.7387 | 1.9084 | 500 | 0.9869 | 72.6842 | 30.8710 |
| 0.3726 | 2.2901 | 600 | 1.0133 | 71.2298 | 28.8136 |
| 0.3741 | 2.6718 | 700 | 0.9793 | 73.6089 | 35.0299 |
Framework versions
- Transformers 4.57.3
- Pytorch 2.9.0+cu126
- Datasets 2.21.0
- Tokenizers 0.22.2