wav2vec2-xlsr-53-CV-demo-google-colab-Ezra_William_Prod19
This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on the common_voice_13_0 dataset. It achieves the following results on the evaluation set:
- Loss: 0.3817
- Wer: 0.3751
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: 16
- eval_batch_size: 8
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 9
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 2.9461 | 1.0 | 278 | 2.9213 | 1.0 |
| 2.8707 | 2.0 | 556 | 2.7842 | 1.0 |
| 1.2835 | 3.0 | 834 | 0.6652 | 0.6095 |
| 0.731 | 4.0 | 1112 | 0.4922 | 0.4826 |
| 0.6299 | 5.0 | 1390 | 0.4429 | 0.4414 |
| 0.5139 | 6.0 | 1668 | 0.4123 | 0.4020 |
| 0.4785 | 7.0 | 1946 | 0.3932 | 0.3874 |
| 0.4636 | 8.0 | 2224 | 0.3895 | 0.3798 |
| 0.4381 | 9.0 | 2502 | 0.3817 | 0.3751 |
Framework versions
- Transformers 4.40.1
- Pytorch 2.3.0+cu121
- Datasets 2.19.0
- Tokenizers 0.19.1