wav2vec2-xls-r-300m-ja-phoneme_cv_14_4
This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the audiofolder dataset. It achieves the following results on the evaluation set:
- Loss: 0.3806
- Wer: 0.0875
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: 2
- eval_batch_size: 1
- seed: 42
- gradient_accumulation_steps: 4
- total_train_batch_size: 8
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- num_epochs: 1
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 5.8739 | 0.1 | 400 | 2.8760 | 1.0 |
| 1.8658 | 0.2 | 800 | 0.7874 | 0.1632 |
| 0.7464 | 0.29 | 1200 | 0.5934 | 0.1300 |
| 0.6031 | 0.39 | 1600 | 0.5022 | 0.1131 |
| 0.5313 | 0.49 | 2000 | 0.4730 | 0.1053 |
| 0.4973 | 0.59 | 2400 | 0.4571 | 0.1012 |
| 0.4686 | 0.69 | 2800 | 0.4156 | 0.0962 |
| 0.4315 | 0.79 | 3200 | 0.3926 | 0.0916 |
| 0.4192 | 0.88 | 3600 | 0.3865 | 0.0886 |
| 0.4055 | 0.98 | 4000 | 0.3806 | 0.0875 |
Framework versions
- Transformers 4.31.0
- Pytorch 2.0.1+cu117
- Datasets 2.14.3
- Tokenizers 0.13.3