wav2vec2-base-arabic-without-reused-en-token-ids
This model is a fine-tuned version of facebook/wav2vec2-base on the common_voice_17_0 dataset. It achieves the following results on the evaluation set:
- Loss: 0.8007
- Wer: 0.6758
- Cer: 0.2119
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 |
|---|---|---|---|---|---|
| 2.9758 | 1.0 | 884 | 3.0186 | 1.0 | 1.0 |
| 1.5795 | 2.0 | 1768 | 1.5033 | 0.9511 | 0.3800 |
| 1.0133 | 3.0 | 2652 | 0.9663 | 0.7720 | 0.2494 |
| 0.8316 | 4.0 | 3536 | 0.8007 | 0.6758 | 0.2119 |
Framework versions
- Transformers 4.52.4
- Pytorch 2.6.0+cu124
- Datasets 2.18.0
- Tokenizers 0.21.2