Whisper Small Tr - CV 43h - Linear Decay
This model is a fine-tuned version of openai/whisper-small on the Common Voice 16.1 dataset. It achieves the following results on the evaluation set:
- Loss: 0.2554
- Wer: 21.6318
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: 16
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 32
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 4000
- training_steps: 4000
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 0.2413 | 0.37 | 500 | 0.2845 | 23.9500 |
| 0.2005 | 0.73 | 1000 | 0.2615 | 22.3219 |
| 0.1583 | 1.1 | 1500 | 0.2493 | 21.5365 |
| 0.1513 | 1.46 | 2000 | 0.2485 | 21.5455 |
| 0.1502 | 1.83 | 2500 | 0.2407 | 21.0639 |
| 0.0959 | 2.19 | 3000 | 0.2485 | 21.1987 |
| 0.1074 | 2.56 | 3500 | 0.2539 | 21.8762 |
| 0.105 | 2.92 | 4000 | 0.2554 | 21.6318 |
Framework versions
- Transformers 4.39.3
- Pytorch 2.2.1+cu121
- Datasets 2.18.0
- Tokenizers 0.15.2