xlsr53-mupe-1
This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53-portuguese on the None dataset. It achieves the following results on the evaluation set:
- Loss: 1.5001
- Wer: 0.5465
- Cer: 0.3049
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: 1e-05
- train_batch_size: 2
- eval_batch_size: 2
- seed: 42
- optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 100
- num_epochs: 12
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
|---|---|---|---|---|---|
| 1.3239 | 1.0 | 2000 | 1.5103 | 0.6366 | 0.3503 |
| 1.156 | 2.0 | 4000 | 1.4288 | 0.6022 | 0.3261 |
| 0.958 | 3.0 | 6000 | 1.4058 | 0.5893 | 0.3214 |
| 1.2899 | 4.0 | 8000 | 1.4745 | 0.5743 | 0.3122 |
| 0.856 | 5.0 | 10000 | 1.4086 | 0.5684 | 0.3195 |
| 0.9923 | 6.0 | 12000 | 1.4499 | 0.5651 | 0.3086 |
| 0.9734 | 7.0 | 14000 | 1.4358 | 0.5579 | 0.3089 |
| 1.084 | 8.0 | 16000 | 1.5082 | 0.5507 | 0.3036 |
| 1.0326 | 9.0 | 18000 | 1.4677 | 0.5579 | 0.3064 |
| 1.229 | 10.0 | 20000 | 1.4917 | 0.5480 | 0.3056 |
| 0.785 | 11.0 | 22000 | 1.4971 | 0.5471 | 0.3050 |
| 0.6886 | 12.0 | 24000 | 1.5001 | 0.5465 | 0.3048 |
Framework versions
- Transformers 4.55.2
- Pytorch 2.7.0+cu126
- Datasets 4.0.0
- Tokenizers 0.21.4