wav2vec2-xlsr-53-CV-demo-google-colab-Ezra_William_Prod17
This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the common_voice_13_0 dataset. It achieves the following results on the evaluation set:
- Loss: 0.3179
- Wer: 0.2729
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 |
|---|---|---|---|---|
| 2.9181 | 1.0 | 278 | 2.8192 | 1.0 |
| 0.8936 | 2.0 | 556 | 0.6359 | 0.6174 |
| 0.4465 | 3.0 | 834 | 0.4092 | 0.4352 |
| 0.3093 | 4.0 | 1112 | 0.3628 | 0.3739 |
| 0.2337 | 5.0 | 1390 | 0.3286 | 0.3354 |
| 0.1831 | 6.0 | 1668 | 0.3377 | 0.3134 |
| 0.1504 | 7.0 | 1946 | 0.3300 | 0.3019 |
| 0.1331 | 8.0 | 2224 | 0.3151 | 0.2978 |
| 0.1203 | 9.0 | 2502 | 0.3023 | 0.2786 |
| 0.107 | 10.0 | 2780 | 0.3157 | 0.2811 |
| 0.095 | 11.0 | 3058 | 0.3116 | 0.2749 |
| 0.0914 | 12.0 | 3336 | 0.3179 | 0.2729 |
Framework versions
- Transformers 4.40.1
- Pytorch 2.3.0+cu121
- Datasets 2.19.0
- Tokenizers 0.19.1