wav2vec2-large-xlsr-53_full_train
This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on the Swissdial dataset. It achieves the following results on the evaluation set:
- Loss: 0.2811
- Wer: 0.2909
Model description
Wav2Vec2-XLSR-53 trained on augmented Swiss Dial dataset
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: 0.0001
- train_batch_size: 16
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 16
- total_train_batch_size: 256
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 1000
- num_epochs: 10
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 1.7666 | 2.69 | 1000 | 0.4356 | 0.4954 |
| 0.7868 | 5.39 | 2000 | 0.2693 | 0.3180 |
| 0.6948 | 8.09 | 3000 | 0.2811 | 0.2909 |
Framework versions
- Transformers 4.19.2
- Pytorch 1.11.0+cu102
- Datasets 2.2.1
- Tokenizers 0.12.1