afrivoices-whisper-turbo
This model is a fine-tuned version of openai/whisper-large-v3-turbo on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.5509
- Wer: 0.2632
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 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: 200
- num_epochs: 3
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 2.1024 | 0.1210 | 1000 | 1.0739 | 0.4950 |
| 1.7731 | 0.2419 | 2000 | 0.9169 | 0.4287 |
| 1.6709 | 0.3629 | 3000 | 0.8682 | 0.3925 |
| 1.5405 | 0.4838 | 4000 | 0.8181 | 0.3641 |
| 1.4395 | 0.6048 | 5000 | 0.7728 | 0.3398 |
| 1.4374 | 0.7257 | 6000 | 0.7291 | 0.3474 |
| 1.3565 | 0.8467 | 7000 | 0.7040 | 0.3477 |
| 1.3255 | 0.9676 | 8000 | 0.6845 | 0.3080 |
| 1.1642 | 1.0885 | 9000 | 0.6755 | 0.2991 |
| 1.1368 | 1.2095 | 10000 | 0.6628 | 0.2991 |
| 1.1517 | 1.3305 | 11000 | 0.6355 | 0.3511 |
| 1.1455 | 1.4514 | 12000 | 0.6197 | 0.2790 |
| 1.1246 | 1.5724 | 13000 | 0.6074 | 0.3015 |
| 1.1126 | 1.6933 | 14000 | 0.5968 | 0.2889 |
| 1.1092 | 1.8143 | 15000 | 0.5874 | 0.3118 |
| 1.0218 | 1.9352 | 16000 | 0.5850 | 0.2715 |
| 0.9080 | 2.0561 | 17000 | 0.5868 | 0.2981 |
| 0.8189 | 2.1771 | 18000 | 0.5804 | 0.2786 |
| 0.7982 | 2.2980 | 19000 | 0.5637 | 0.2824 |
| 0.8133 | 2.4190 | 20000 | 0.5823 | 0.2711 |
| 0.8397 | 2.5399 | 21000 | 0.5702 | 0.2773 |
| 0.8672 | 2.6609 | 22000 | 0.5611 | 0.2721 |
| 0.8279 | 2.7819 | 23000 | 0.5568 | 0.2632 |
| 0.8361 | 2.9028 | 24000 | 0.5495 | 0.2851 |
| 0.7827 | 3.0 | 24804 | 0.5509 | 0.2632 |
Framework versions
- Transformers 5.12.1
- Pytorch 2.11.0+cu128
- Datasets 5.0.0
- Tokenizers 0.22.2