wav2vec2-large-mms-1b-yoruba-colab
This model is a fine-tuned version of facebook/mms-1b-all on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.6193
- Wer: 0.5143
- Cer: 0.1780
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: 0.001
- train_batch_size: 16
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 32
- 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: 100
- training_steps: 2000
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
|---|---|---|---|---|---|
| 1.7403 | 0.125 | 250 | 0.6963 | 0.5774 | 0.1984 |
| 1.6613 | 1.0445 | 500 | 0.6728 | 0.5636 | 0.2001 |
| 1.6135 | 1.1695 | 750 | 0.6625 | 0.5486 | 0.1919 |
| 1.5472 | 2.089 | 1000 | 0.6457 | 0.5448 | 0.1898 |
| 1.5304 | 3.0085 | 1250 | 0.6439 | 0.5370 | 0.1877 |
| 1.5243 | 3.1335 | 1500 | 0.6335 | 0.5179 | 0.1811 |
| 1.4798 | 4.053 | 1750 | 0.6264 | 0.5181 | 0.1803 |
| 1.4342 | 4.178 | 2000 | 0.6193 | 0.5143 | 0.1780 |
Framework versions
- Transformers 5.13.1
- Pytorch 2.11.0+cu128
- Datasets 4.0.0
- Tokenizers 0.22.2