<!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. -->
wav2vec2-large-arabic
This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on the common_voice dataset. It achieves the following results on the evaluation set:
- Loss: 0.7193
- Wer: 0.5399
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: 0.0003
- train_batch_size: 8
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 4
- total_train_batch_size: 32
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- num_epochs: 5
- mixed_precision_training: Native AMP
Training results
Training Loss | Epoch | Step | Validation Loss | Wer |
---|---|---|---|---|
2.0113 | 0.57 | 500 | 1.4827 | 0.9062 |
0.8047 | 1.14 | 1000 | 0.7848 | 0.6483 |
0.6679 | 1.71 | 1500 | 0.6610 | 0.5899 |
0.5226 | 2.28 | 2000 | 0.6689 | 0.5877 |
0.4968 | 2.85 | 2500 | 0.6567 | 0.5847 |
0.3756 | 3.42 | 3000 | 0.5517 | 0.5093 |
0.3464 | 3.99 | 3500 | 0.5855 | 0.5147 |
0.2662 | 4.57 | 4000 | 0.7193 | 0.5399 |
Framework versions
- Transformers 4.20.1
- Pytorch 1.10.0+cu113
- Datasets 2.1.0
- Tokenizers 0.12.1