wav2vec2-large-xls-r-300m-turkish-colab
This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the common_voice dataset. It achieves the following results on the evaluation set:
- Loss: 0.3821
- Wer: 0.3208
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: Adam with betas=(0.9,0.999) and epsilon=1e-08
- 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 |
|---|---|---|---|---|
| 3.9162 | 3.67 | 400 | 0.6340 | 0.6360 |
| 0.4033 | 7.34 | 800 | 0.4588 | 0.4911 |
| 0.1919 | 11.01 | 1200 | 0.4392 | 0.4460 |
| 0.1315 | 14.68 | 1600 | 0.4269 | 0.4270 |
| 0.0963 | 18.35 | 2000 | 0.4327 | 0.3834 |
| 0.0801 | 22.02 | 2400 | 0.3867 | 0.3643 |
| 0.0631 | 25.69 | 2800 | 0.3854 | 0.3441 |
| 0.0492 | 29.36 | 3200 | 0.3821 | 0.3208 |
Framework versions
- Transformers 4.11.3
- Pytorch 1.10.0+cu113
- Datasets 1.18.3
- Tokenizers 0.10.3