2nd-wav2vec2-l-xls-r-300m-turkish-test
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: 0.6019
- Wer: 0.4444
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: 0.001
- 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: 20
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 3.0522 | 3.67 | 400 | 0.7773 | 0.7296 |
| 0.5369 | 7.34 | 800 | 0.6282 | 0.5888 |
| 0.276 | 11.01 | 1200 | 0.5998 | 0.5330 |
| 0.1725 | 14.68 | 1600 | 0.5859 | 0.4908 |
| 0.1177 | 18.35 | 2000 | 0.6019 | 0.4444 |
Framework versions
- Transformers 4.11.3
- Pytorch 1.10.0+cu111
- Datasets 1.18.3
- Tokenizers 0.10.3