<!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. -->
b31-wav2vec2-large-xls-r-romansh-colab
This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the common_voice_13_0 dataset. It achieves the following results on the evaluation set:
- Loss: 0.4908
- Wer: 0.3386
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: 4
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 8
- 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 |
|---|---|---|---|---|
| 4.8062 | 3.05 | 400 | 2.9021 | 1.0 |
| 1.1417 | 6.11 | 800 | 0.5258 | 0.5452 |
| 0.269 | 9.16 | 1200 | 0.4839 | 0.4565 |
| 0.1567 | 12.21 | 1600 | 0.5115 | 0.4194 |
| 0.1163 | 15.27 | 2000 | 0.5361 | 0.4199 |
| 0.089 | 18.32 | 2400 | 0.4787 | 0.3852 |
| 0.0713 | 21.37 | 2800 | 0.5061 | 0.3701 |
| 0.0565 | 24.43 | 3200 | 0.5111 | 0.3535 |
| 0.0456 | 27.48 | 3600 | 0.4908 | 0.3386 |
Framework versions
- Transformers 4.26.0
- Pytorch 2.0.1+cu118
- Datasets 2.14.4
- Tokenizers 0.13.3
