wolof
This model is a fine-tuned version of LeBenchmark/wav2vec2-FR-2.6K-base on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.2816
- Wer: 0.3897
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: 20
- 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: 10
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 2.9468 | 1.67 | 1500 | 0.7036 | 0.6418 |
| 0.5506 | 3.33 | 3000 | 0.4129 | 0.5018 |
| 0.3817 | 5.0 | 4500 | 0.3414 | 0.4519 |
| 0.2885 | 6.67 | 6000 | 0.3181 | 0.4305 |
| 0.2275 | 8.33 | 7500 | 0.2920 | 0.4011 |
| 0.1852 | 10.0 | 9000 | 0.2816 | 0.3897 |
Framework versions
- Transformers 4.18.0
- Pytorch 1.11.0+cu102
- Datasets 2.1.0
- Tokenizers 0.12.1