Whisper Medium Tr - denysdios
This model is a fine-tuned version of openai/whisper-medium on the Common Voice 13.0 & Fleurs dataset. It achieves the following results on the evaluation set:
- Loss: 0.1618
- Wer: 14.3825
Model description
The model took about nine hours to train on a single A100 GPU.
Intended uses & limitations
Absolutely no restrictions additional to whisper models. Increasing the Turkish labeled data in whisper, which was 4333/690k (0.0063), was the primary objective. There are just 49.945 hours of data in the fine-tuning dataset, or about 1.1% of the Turkish dataset that has already been trained.
Training and evaluation data
Processing...
Training procedure
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 1e-05
- train_batch_size: 16
- eval_batch_size: 16
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- training_steps: 4000
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 0.1803 | 0.36 | 1000 | 0.2089 | 18.6326 |
| 0.1428 | 0.71 | 2000 | 0.1821 | 16.3912 |
| 0.0535 | 1.07 | 3000 | 0.1693 | 14.9132 |
| 0.0491 | 1.43 | 4000 | 0.1618 | 14.3825 |
Framework versions
- Transformers 4.38.1
- Pytorch 2.1.0+cu121
- Datasets 2.17.1
- Tokenizers 0.15.2
- Downloads last month
- 0