Whisper Small Ar Tashkeel - AzeemX
This model is a fine-tuned version of arbml/whisper-small-ar on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.0548
- Wer: 80.0678
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: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- training_steps: 7000
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 0.1104 | 1.0320 | 1000 | 0.1108 | 95.5495 |
| 0.042 | 2.0640 | 2000 | 0.0791 | 60.0267 |
| 0.015 | 3.0960 | 3000 | 0.0622 | 61.8399 |
| 0.0051 | 4.1280 | 4000 | 0.0548 | 48.6347 |
| 0.0021 | 5.1600 | 5000 | 0.0521 | 67.4069 |
| 0.0007 | 6.1920 | 6000 | 0.0539 | 71.4779 |
| 0.0004 | 7.2239 | 7000 | 0.0548 | 80.0678 |
Framework versions
- Transformers 4.45.2
- Pytorch 2.3.1
- Datasets 3.0.1
- Tokenizers 0.20.0