dgx1_w2v2_base_finetune_teacher_babble_noise_libri_360_hours_50_epochs_batch_4
This model is a fine-tuned version of facebook/wav2vec2-base-960h on the None dataset. It achieves the following results on the evaluation set:
- Loss: 48.7078
- Wer: 0.2642
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: 0.0005
- train_batch_size: 4
- eval_batch_size: 1
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_ratio: 0.2
- num_epochs: 1
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 484.5509 | 0.02 | 500 | 28.4641 | 0.1563 |
| 293.5338 | 0.04 | 1000 | 25.8292 | 0.1446 |
| 293.9193 | 0.06 | 1500 | 29.0716 | 0.1710 |
| 322.8119 | 0.08 | 2000 | 34.0894 | 0.1971 |
| 365.3664 | 0.1 | 2500 | 36.7096 | 0.2180 |
| 404.8819 | 0.12 | 3000 | 41.7040 | 0.2413 |
| 435.5897 | 0.13 | 3500 | 48.7078 | 0.2642 |
Framework versions
- Transformers 4.25.1
- Pytorch 1.12.1
- Datasets 2.8.0
- Tokenizers 0.13.2