XLS-R_53_english
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: 0.3430
- Wer: 0.3033
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 | Wer |
|---|---|---|---|---|
| 4.6589 | 1.65 | 500 | 3.1548 | 1.0 |
| 2.5363 | 3.3 | 1000 | 1.0250 | 0.8707 |
| 0.849 | 4.95 | 1500 | 0.3964 | 0.4636 |
| 0.4812 | 6.6 | 2000 | 0.3341 | 0.3907 |
| 0.3471 | 8.25 | 2500 | 0.3351 | 0.3659 |
| 0.2797 | 9.9 | 3000 | 0.3104 | 0.3475 |
| 0.2336 | 11.55 | 3500 | 0.3545 | 0.3419 |
| 0.2116 | 13.2 | 4000 | 0.3577 | 0.3353 |
| 0.1688 | 14.85 | 4500 | 0.3383 | 0.3302 |
| 0.1587 | 16.5 | 5000 | 0.3431 | 0.3235 |
| 0.1358 | 18.15 | 5500 | 0.3504 | 0.3209 |
| 0.1323 | 19.8 | 6000 | 0.3468 | 0.3191 |
| 0.115 | 21.45 | 6500 | 0.3331 | 0.3127 |
| 0.108 | 23.1 | 7000 | 0.3497 | 0.3099 |
| 0.0938 | 24.75 | 7500 | 0.3532 | 0.3091 |
| 0.0974 | 26.4 | 8000 | 0.3461 | 0.3086 |
| 0.0867 | 28.05 | 8500 | 0.3422 | 0.3054 |
| 0.0852 | 29.7 | 9000 | 0.3430 | 0.3033 |
Framework versions
- Transformers 4.17.0
- Pytorch 1.12.1+cu113
- Datasets 1.18.3
- Tokenizers 0.12.1