asr-for-phoneme-detection-1
This model is a fine-tuned version of facebook/wav2vec2-base on the TIMIT dataset. It achieves the following results on the evaluation set:
- Loss: 0.2756
- Wer: 0.9613
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: 4
- 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: 4
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 1.8867 | 0.74 | 500 | 1.8995 | 1.0 |
| 0.9265 | 1.49 | 1000 | 0.5039 | 1.0 |
| 0.4881 | 2.23 | 1500 | 0.3537 | 0.9940 |
| 0.3512 | 2.98 | 2000 | 0.2781 | 0.9702 |
| 0.2706 | 3.72 | 2500 | 0.2756 | 0.9613 |
Framework versions
- Transformers 4.11.3
- Pytorch 2.4.0
- Datasets 2.20.0
- Tokenizers 0.10.3