This model is a fine-tuned version of openai/whisper-base.en on the None dataset. It achieves the following results on the evaluation set:
More information needed
More information needed
For training and evaluation, we use a Pakistani English accent dataset to improve the model’s robustness to Pakistani English accents and minimize the Word Error Rate (WER).
The following hyperparameters were used during training:
| Training Loss | Epoch | Step | Validation Loss | Wer Ortho | Wer |
|---|---|---|---|---|---|
| 0.7801 | 0.8065 | 100 | 0.3812 | 13.3555 | 9.4194 |
| 0.2817 | 1.6129 | 200 | 0.3647 | 12.5674 | 9.1352 |
| 0.1853 | 2.4194 | 300 | 0.3705 | 12.7748 | 8.8510 |
| 0.146 | 3.2258 | 400 | 0.3738 | 12.5674 | 9.0946 |
| 0.1097 | 4.0323 | 500 | 0.3797 | 12.8163 | 9.1352 |
model.safetensors 145 MBtraining_args.bin 5.4 kBopenai/whisper-base.en