wav2vec2-large-xls-r-300m-ar-colab
This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the common_voice_6_0 dataset. It achieves the following results on the evaluation set:
- Loss: 0.9769
- Wer: 0.5556
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: 16
- eval_batch_size: 8
- 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: linear
- lr_scheduler_warmup_steps: 500
- num_epochs: 30
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 5.7757 | 4.4944 | 400 | 3.1422 | 1.0 |
| 1.4911 | 8.9888 | 800 | 0.8969 | 0.6592 |
| 0.4137 | 13.4831 | 1200 | 1.1337 | 0.7046 |
| 0.2215 | 17.9775 | 1600 | 0.9063 | 0.5733 |
| 0.135 | 22.4719 | 2000 | 0.9463 | 0.5761 |
| 0.0989 | 26.9663 | 2400 | 0.9769 | 0.5556 |
Framework versions
- Transformers 4.52.4
- Pytorch 2.6.0+cu124
- Datasets 2.18.0
- Tokenizers 0.21.2