wav2vec2-xls-r-timit-tokenizer-base
This model is a fine-tuned version of facebook/wav2vec2-base on the None dataset. It achieves the following results on the evaluation set:
- Loss: 3.0828
- Wer: 1.0
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 |
|---|---|---|---|---|
| 3.3134 | 4.03 | 500 | 3.0814 | 1.0 |
| 2.9668 | 8.06 | 1000 | 3.0437 | 1.0 |
| 2.9604 | 12.1 | 1500 | 3.0337 | 1.0 |
| 2.9619 | 16.13 | 2000 | 3.0487 | 1.0 |
| 2.9588 | 20.16 | 2500 | 3.0859 | 1.0 |
| 2.957 | 24.19 | 3000 | 3.0921 | 1.0 |
| 2.9555 | 28.22 | 3500 | 3.0828 | 1.0 |
Framework versions
- Transformers 4.11.3
- Pytorch 1.10.0+cu111
- Datasets 1.13.3
- Tokenizers 0.10.3