<!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. -->
wav2vec2-base-timit-demo-google-colab
This model is a fine-tuned version of facebook/wav2vec2-base on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.5366
- Wer: 0.3452
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: 16
- 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: 40
- mixed_precision_training: Native AMP
Training results
Training Loss | Epoch | Step | Validation Loss | Wer |
---|---|---|---|---|
3.5499 | 2.01 | 500 | 1.9780 | 0.9933 |
0.7517 | 4.02 | 1000 | 0.4654 | 0.4720 |
0.2953 | 6.02 | 1500 | 0.4202 | 0.4049 |
0.1809 | 8.03 | 2000 | 0.4276 | 0.3759 |
0.1335 | 10.04 | 2500 | 0.4458 | 0.3774 |
0.107 | 12.05 | 3000 | 0.4559 | 0.3707 |
0.0923 | 14.06 | 3500 | 0.4607 | 0.3659 |
0.0753 | 16.06 | 4000 | 0.4699 | 0.3531 |
0.0658 | 18.07 | 4500 | 0.4507 | 0.3588 |
0.0569 | 20.08 | 5000 | 0.5089 | 0.3532 |
0.0493 | 22.09 | 5500 | 0.5481 | 0.3515 |
0.043 | 24.1 | 6000 | 0.5066 | 0.3528 |
0.0388 | 26.1 | 6500 | 0.5418 | 0.3534 |
0.034 | 28.11 | 7000 | 0.5566 | 0.3524 |
0.03 | 30.12 | 7500 | 0.4994 | 0.3437 |
0.0274 | 32.13 | 8000 | 0.5588 | 0.3520 |
0.0239 | 34.14 | 8500 | 0.5328 | 0.3458 |
0.0212 | 36.14 | 9000 | 0.5221 | 0.3467 |
0.0186 | 38.15 | 9500 | 0.5366 | 0.3452 |
Framework versions
- Transformers 4.17.0
- Pytorch 1.12.0+cu113
- Datasets 1.18.3
- Tokenizers 0.12.1