wav2vec2-xlsr-afrispeech-all
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: 0.3638
- Model Preparation Time: 0.0027
- Wer: 0.266
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: 3589
- num_epochs: 10
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Model Preparation Time | Wer |
|---|---|---|---|---|---|
| 1.2714 | 1.0 | 3590 | 0.5052 | 0.0027 | 0.356 |
| 1.2401 | 2.0 | 7180 | 0.4329 | 0.0027 | 0.326 |
| 0.8673 | 3.0 | 10770 | 0.3992 | 0.0027 | 0.308 |
| 0.7546 | 4.0 | 14360 | 0.3905 | 0.0027 | 0.297 |
| 0.7386 | 5.0 | 17950 | 0.3879 | 0.0027 | 0.286 |
| 0.6193 | 6.0 | 21540 | 0.3751 | 0.0027 | 0.281 |
| 0.5586 | 7.0 | 25130 | 0.3699 | 0.0027 | 0.276 |
| 0.5194 | 8.0 | 28720 | 0.3711 | 0.0027 | 0.273 |
| 0.4453 | 9.0 | 32310 | 0.3636 | 0.0027 | 0.267 |
| 0.4618 | 10.0 | 35900 | 0.3638 | 0.0027 | 0.266 |
Framework versions
- Transformers 5.3.0
- Pytorch 2.12.0.dev20260307+cu128
- Datasets 2.19.0
- Tokenizers 0.22.2