wav2vec2-xlsr-dora-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: 0.4356
- Wer: 0.282
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 | Wer |
|---|---|---|---|---|
| 1.2810 | 1.0 | 1350 | 0.5821 | 0.377 |
| 1.1303 | 2.0 | 2700 | 0.5169 | 0.329 |
| 0.9971 | 3.0 | 4050 | 0.4873 | 0.313 |
| 1.0113 | 4.0 | 5400 | 0.4681 | 0.301 |
| 0.9989 | 5.0 | 6750 | 0.4578 | 0.296 |
| 1.0076 | 6.0 | 8100 | 0.4511 | 0.293 |
| 0.8525 | 7.0 | 9450 | 0.4426 | 0.286 |
| 0.9018 | 8.0 | 10800 | 0.4387 | 0.282 |
| 0.8502 | 9.0 | 12150 | 0.4378 | 0.281 |
| 0.8966 | 10.0 | 13500 | 0.4356 | 0.282 |
Framework versions
- PEFT 0.18.1
- Transformers 5.3.0
- Pytorch 2.12.0.dev20260307+cu128
- Datasets 2.19.0
- Tokenizers 0.22.2