480min_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: 0.9643
- Wer: 53.2248
- Cer: 19.2965
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: 2
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 16
- total_train_batch_size: 32
- 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: 3000
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
|---|---|---|---|---|---|
| 1.4334 | 0.7630 | 100 | 1.0494 | 72.2732 | 28.1859 |
| 0.9941 | 1.5188 | 200 | 0.8512 | 63.5631 | 24.9170 |
| 0.7004 | 2.2747 | 300 | 0.7689 | 60.1091 | 23.0066 |
| 0.6149 | 3.0305 | 400 | 0.7136 | 56.9791 | 23.1324 |
| 0.48 | 3.7935 | 500 | 0.7019 | 56.0939 | 20.9567 |
| 0.2873 | 4.5494 | 600 | 0.7126 | 54.2681 | 20.2168 |
| 0.2662 | 5.3052 | 700 | 0.7403 | 55.5090 | 20.8096 |
| 0.1792 | 6.0610 | 800 | 0.7632 | 54.2128 | 19.5224 |
| 0.1018 | 6.8240 | 900 | 0.7796 | 54.5843 | 20.1046 |
| 0.0706 | 7.5799 | 1000 | 0.8194 | 52.8138 | 18.7567 |
| 0.057 | 8.3357 | 1100 | 0.8403 | 54.4815 | 20.3472 |
| 0.0296 | 9.0916 | 1200 | 0.8704 | 53.5963 | 19.3738 |
| 0.019 | 9.8546 | 1300 | 0.8982 | 54.4341 | 19.9060 |
| 0.0151 | 10.6104 | 1400 | 0.9124 | 53.6279 | 19.5209 |
| 0.0084 | 11.3662 | 1500 | 0.9445 | 53.3750 | 19.1646 |
| 0.0068 | 12.1221 | 1600 | 0.9629 | 53.2011 | 19.2404 |
| 0.0056 | 12.8851 | 1700 | 0.9643 | 53.2248 | 19.2965 |
Framework versions
- Transformers 4.57.3
- Pytorch 2.9.0+cu126
- Datasets 2.21.0
- Tokenizers 0.22.2