wav2vec2-base-arabic-without-reused-en-token-ids-v1
This model is a fine-tuned version of Mohammadawad1/wav2vec2-base-arabic-without-reused-en-token-ids on the common_voice_17_0 dataset. It achieves the following results on the evaluation set:
- Loss: 0.6458
- Wer: 0.5747
- Cer: 0.1746
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: 16
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 32
- optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: constant_with_warmup
- lr_scheduler_warmup_steps: 50
- num_epochs: 4
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
|---|---|---|---|---|---|
| 0.7068 | 1.0 | 884 | 0.7127 | 0.6490 | 0.2005 |
| 0.6227 | 2.0 | 1768 | 0.6717 | 0.6110 | 0.1874 |
| 0.5755 | 3.0 | 2652 | 0.6369 | 0.5888 | 0.1790 |
| 0.5331 | 4.0 | 3536 | 0.6458 | 0.5747 | 0.1746 |
Framework versions
- Transformers 4.52.4
- Pytorch 2.6.0+cu124
- Datasets 2.18.0
- Tokenizers 0.21.2