wav2vec2-xlsr-53-CV-demo-google-colab-Ezra_William_Prod14
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: 0.3505
- Wer: 0.3279
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: 12
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 2.9451 | 1.0 | 278 | 2.9182 | 1.0 |
| 2.87 | 2.0 | 556 | 2.7116 | 1.0 |
| 1.1102 | 3.0 | 834 | 0.6030 | 0.5907 |
| 0.6952 | 4.0 | 1112 | 0.4691 | 0.4755 |
| 0.5976 | 5.0 | 1390 | 0.4316 | 0.4263 |
| 0.4842 | 6.0 | 1668 | 0.3887 | 0.3842 |
| 0.4444 | 7.0 | 1946 | 0.3722 | 0.3670 |
| 0.4221 | 8.0 | 2224 | 0.3721 | 0.3538 |
| 0.3929 | 9.0 | 2502 | 0.3527 | 0.3463 |
| 0.3611 | 10.0 | 2780 | 0.3538 | 0.3386 |
| 0.3669 | 11.0 | 3058 | 0.3513 | 0.3303 |
| 0.3517 | 12.0 | 3336 | 0.3505 | 0.3279 |
Framework versions
- Transformers 4.40.0
- Pytorch 2.2.2+cu121
- Datasets 2.18.0
- Tokenizers 0.19.1