wav2vec2-xlsr-53-torgo-8batch-30epochs-500steps
This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on the None dataset. It achieves the following results on the evaluation set:
- Loss: 1.0747
- Cer: 0.3015
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: 8
- 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 | Cer |
|---|---|---|---|---|
| 25.2161 | 2.53 | 500 | 5.2319 | 0.9816 |
| 3.649 | 5.05 | 1000 | 3.7028 | 0.9816 |
| 2.8241 | 7.58 | 1500 | 2.0426 | 0.8802 |
| 1.6018 | 10.1 | 2000 | 1.2611 | 0.4394 |
| 1.0374 | 12.63 | 2500 | 1.1117 | 0.3713 |
| 0.7962 | 15.15 | 3000 | 1.0730 | 0.3445 |
| 0.6436 | 17.68 | 3500 | 1.1026 | 0.3309 |
| 0.5344 | 20.2 | 4000 | 1.0734 | 0.3100 |
| 0.4556 | 22.73 | 4500 | 1.0877 | 0.3177 |
| 0.3992 | 25.25 | 5000 | 1.0614 | 0.3065 |
| 0.3764 | 27.78 | 5500 | 1.0747 | 0.3015 |
Framework versions
- Transformers 4.18.0
- Pytorch 1.12.1+cu113
- Datasets 1.18.3
- Tokenizers 0.12.1