wav2vec2-xlsr-afrispeech-general
This model is a fine-tuned version of jonatasgrosman/wav2vec2-large-xlsr-53-english on the afrispeech-200 dataset. It achieves the following results on the evaluation set:
- Loss: 0.3592
- Model Preparation Time: 0.0069
- Wer: 0.252
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
- gradient_accumulation_steps: 2
- total_train_batch_size: 16
- 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: 1349
- num_epochs: 10
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Model Preparation Time | Wer |
|---|---|---|---|---|---|
| 1.1859 | 1.0 | 1350 | 0.5450 | 0.0069 | 0.343 |
| 0.8970 | 2.0 | 2700 | 0.4383 | 0.0069 | 0.305 |
| 0.7007 | 3.0 | 4050 | 0.4032 | 0.0069 | 0.288 |
| 0.6468 | 4.0 | 5400 | 0.3804 | 0.0069 | 0.278 |
| 0.5917 | 5.0 | 6750 | 0.3769 | 0.0069 | 0.27 |
| 0.5619 | 6.0 | 8100 | 0.3683 | 0.0069 | 0.264 |
| 0.4550 | 7.0 | 9450 | 0.3668 | 0.0069 | 0.262 |
| 0.4735 | 8.0 | 10800 | 0.3631 | 0.0069 | 0.256 |
| 0.3967 | 9.0 | 12150 | 0.3662 | 0.0069 | 0.253 |
| 0.4371 | 10.0 | 13500 | 0.3592 | 0.0069 | 0.252 |
Framework versions
- Transformers 5.3.0
- Pytorch 2.12.0.dev20260307+cu128
- Datasets 2.19.0
- Tokenizers 0.22.2