wav2vec2-large-timit-demo-google-colab
This model is a fine-tuned version of facebook/wav2vec2-large on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.4603
- Wer: 0.3096
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: 8
- 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: 30
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 3.3271 | 1.0 | 500 | 1.0169 | 0.8540 |
| 0.8682 | 2.01 | 1000 | 0.5415 | 0.5366 |
| 0.6008 | 3.01 | 1500 | 0.5105 | 0.4881 |
| 0.4592 | 4.02 | 2000 | 0.5244 | 0.4596 |
| 0.3753 | 5.02 | 2500 | 0.4762 | 0.4280 |
| 0.3271 | 6.02 | 3000 | 0.4418 | 0.4035 |
| 0.2787 | 7.03 | 3500 | 0.5069 | 0.4033 |
| 0.2594 | 8.03 | 4000 | 0.5346 | 0.3871 |
| 0.2253 | 9.04 | 4500 | 0.5057 | 0.3847 |
| 0.2133 | 10.04 | 5000 | 0.5521 | 0.3772 |
| 0.1889 | 11.04 | 5500 | 0.4800 | 0.3697 |
| 0.1694 | 12.05 | 6000 | 0.5778 | 0.3749 |
| 0.1571 | 13.05 | 6500 | 0.5368 | 0.3669 |
| 0.1483 | 14.06 | 7000 | 0.5144 | 0.3577 |
| 0.141 | 15.06 | 7500 | 0.5838 | 0.3589 |
| 0.1238 | 16.06 | 8000 | 0.5242 | 0.3657 |
| 0.1138 | 17.07 | 8500 | 0.5712 | 0.3579 |
| 0.1059 | 18.07 | 9000 | 0.5715 | 0.3495 |
| 0.0994 | 19.08 | 9500 | 0.4920 | 0.3376 |
| 0.0899 | 20.08 | 10000 | 0.4696 | 0.3336 |
| 0.0795 | 21.08 | 10500 | 0.5512 | 0.3278 |
| 0.0759 | 22.09 | 11000 | 0.5155 | 0.3218 |
| 0.0722 | 23.09 | 11500 | 0.4937 | 0.3175 |
| 0.0625 | 24.1 | 12000 | 0.5750 | 0.3245 |
| 0.0579 | 25.1 | 12500 | 0.5473 | 0.3200 |
| 0.0549 | 26.1 | 13000 | 0.5079 | 0.3145 |
| 0.0463 | 27.11 | 13500 | 0.4895 | 0.3164 |
| 0.0457 | 28.11 | 14000 | 0.4757 | 0.3117 |
| 0.0481 | 29.12 | 14500 | 0.4603 | 0.3096 |
Framework versions
- Transformers 4.17.0
- Pytorch 1.12.0+cu113
- Datasets 1.18.3
- Tokenizers 0.12.1