SpeechT5 STT Wav2Vec2
This model is a fine-tuned version of facebook/wav2vec2-base-960h on the Lj-Speech dataset. It achieves the following results on the evaluation set:
- Loss: 252.7729
- Wer: 1.0
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
- gradient_accumulation_steps: 4
- total_train_batch_size: 16
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 100
- num_epochs: 5
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 130.9264 | 0.5319 | 50 | 256.5228 | 0.9827 |
| 134.1007 | 1.0638 | 100 | 256.2832 | 0.9827 |
| 131.0841 | 1.5957 | 150 | 253.9561 | 0.9827 |
| 132.4283 | 2.1277 | 200 | 254.4677 | 0.9827 |
| 137.3693 | 2.6596 | 250 | 254.6855 | 1.0 |
| 128.1369 | 3.1915 | 300 | 252.8348 | 1.0 |
| 132.3826 | 3.7234 | 350 | 254.7122 | 1.0 |
| 130.9401 | 4.2553 | 400 | 254.6629 | 1.0 |
| 129.5693 | 4.7872 | 450 | 252.7729 | 1.0 |
Framework versions
- Transformers 4.42.3
- Pytorch 2.3.0+cu121
- Datasets 2.20.0
- Tokenizers 0.19.1
- Downloads last month
- 1