whisper-small-E30_Yspeed
This model is a fine-tuned version of openai/whisper-small on the aihub old adult freq speed pause changed dataset. It achieves the following results on the evaluation set:
- Loss: 0.1709
- Cer: 4.9342
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: 8
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 16
- optimizer: Use adamw_torch 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: 50
- num_epochs: 2
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Cer |
|---|---|---|---|---|
| 0.2949 | 0.1289 | 100 | 0.2524 | 6.4673 |
| 0.1704 | 0.2579 | 200 | 0.2182 | 5.8741 |
| 0.1803 | 0.3868 | 300 | 0.2094 | 5.7096 |
| 0.1582 | 0.5158 | 400 | 0.1961 | 5.3278 |
| 0.1317 | 0.6447 | 500 | 0.1973 | 5.4805 |
| 0.1237 | 0.7737 | 600 | 0.1821 | 5.1809 |
| 0.1179 | 0.9026 | 700 | 0.1788 | 5.0693 |
| 0.0689 | 1.0309 | 800 | 0.1738 | 4.8696 |
| 0.0543 | 1.1599 | 900 | 0.1783 | 5.1457 |
| 0.0568 | 1.2888 | 1000 | 0.1741 | 5.0458 |
| 0.0552 | 1.4178 | 1100 | 0.1749 | 4.9988 |
| 0.0491 | 1.5467 | 1200 | 0.1724 | 4.8520 |
| 0.0507 | 1.6757 | 1300 | 0.1718 | 5.0106 |
| 0.0429 | 1.8046 | 1400 | 0.1708 | 4.9812 |
| 0.0474 | 1.9336 | 1500 | 0.1709 | 4.9342 |
Framework versions
- Transformers 4.47.0.dev0
- Pytorch 2.5.0+cu121
- Datasets 3.1.0
- Tokenizers 0.20.3