wav2vec2-base-timit-demo-colab
This model is a fine-tuned version of facebook/wav2vec2-base on the TIMIT dataset. It achieves the following results on the evaluation set:
- Loss: 0.4491
- Wer: 0.3382
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: 0.0001
- train_batch_size: 32
- eval_batch_size: 8
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 1000
- num_epochs: 30
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 3.4787 | 4.0 | 500 | 1.4190 | 0.9939 |
| 0.5835 | 8.0 | 1000 | 0.4711 | 0.4370 |
| 0.219 | 12.0 | 1500 | 0.4555 | 0.3994 |
| 0.1251 | 16.0 | 2000 | 0.4515 | 0.3654 |
| 0.0834 | 20.0 | 2500 | 0.4923 | 0.3564 |
| 0.0632 | 24.0 | 3000 | 0.4410 | 0.3399 |
| 0.0491 | 28.0 | 3500 | 0.4491 | 0.3382 |
Framework versions
- Transformers 4.11.3
- Pytorch 1.10.0+cu111
- Datasets 1.18.3
- Tokenizers 0.10.3