This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the MOZILLA-FOUNDATION/COMMON_VOICE_7_0 - ET dataset. It achieves the following results on the evaluation set:
- Loss: 0.4835
- Wer: 0.3475
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: 72
- eval_batch_size: 72
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 144
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_steps: 500
- num_epochs: 100.0
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 0.3825 | 12.5 | 500 | 0.4022 | 0.5059 |
| 0.1592 | 25.0 | 1000 | 0.4585 | 0.4456 |
| 0.1215 | 37.5 | 1500 | 0.4550 | 0.4164 |
| 0.0972 | 50.0 | 2000 | 0.4725 | 0.4088 |
| 0.0731 | 62.5 | 2500 | 0.4568 | 0.3824 |
| 0.0527 | 75.0 | 3000 | 0.4712 | 0.3653 |
| 0.0428 | 87.5 | 3500 | 0.4813 | 0.3520 |
| 0.0383 | 100.0 | 4000 | 0.4835 | 0.3475 |
Framework versions
- Transformers 4.16.0.dev0
- Pytorch 1.10.1+cu102
- Datasets 1.18.4.dev0
- Tokenizers 0.11.0