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: 0.3863
- Wer: 0.3095
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 |
|---|---|---|---|---|
| 3.8284 | 3.67 | 400 | 0.6782 | 0.6739 |
| 0.4174 | 7.34 | 800 | 0.4524 | 0.4811 |
| 0.2015 | 11.01 | 1200 | 0.4736 | 0.4311 |
| 0.1371 | 14.68 | 1600 | 0.4254 | 0.3929 |
| 0.0997 | 18.35 | 2000 | 0.4254 | 0.3636 |
| 0.082 | 22.02 | 2400 | 0.3807 | 0.3474 |
| 0.0665 | 25.69 | 2800 | 0.3987 | 0.3236 |
| 0.0523 | 29.36 | 3200 | 0.3863 | 0.3095 |
Framework versions
- Transformers 4.11.3
- Pytorch 1.10.0+cu111
- Datasets 1.18.3
- Tokenizers 0.10.3