wav2vec2-l2-arctic-phoneme-model
This model is a fine-tuned version of facebook/wav2vec2-base on the None dataset. It achieves the following results on the evaluation set:
- Loss: 284.0506
- Per: 0.1047
- F1: 0.9142
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: 5e-05
- train_batch_size: 16
- eval_batch_size: 16
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 32
- optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED 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: 104
- num_epochs: 10
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Per | F1 |
|---|---|---|---|---|---|
| 10769.5240 | 1.0 | 104 | 1769.2606 | 1.0 | 0.0 |
| 3577.9186 | 2.0 | 208 | 1750.9899 | 1.0 | 0.0 |
| 3495.1671 | 3.0 | 312 | 1570.0214 | 1.0 | 0.0 |
| 2535.6373 | 4.0 | 416 | 783.3173 | 0.3755 | 0.7214 |
| 1357.6378 | 5.0 | 520 | 455.1570 | 0.1670 | 0.8635 |
| 928.3119 | 6.0 | 624 | 352.7345 | 0.1363 | 0.8895 |
| 750.2390 | 7.0 | 728 | 316.2893 | 0.1147 | 0.9068 |
| 650.8412 | 8.0 | 832 | 296.6050 | 0.1112 | 0.9090 |
| 596.2695 | 9.0 | 936 | 287.3040 | 0.1067 | 0.9131 |
| 569.7062 | 10.0 | 1040 | 284.0506 | 0.1047 | 0.9142 |
Framework versions
- Transformers 5.14.1
- Pytorch 2.13.0+cu130
- Datasets 5.0.1
- Tokenizers 0.22.2