<!-- 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-custom-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: 0.7785
- Wer: 0.3534
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: 1
- eval_batch_size: 1
- 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: 10
- mixed_precision_training: Native AMP
Training results
Training Loss | Epoch | Step | Validation Loss | Wer |
---|---|---|---|---|
0.4783 | 0.3 | 500 | 0.7199 | 0.5564 |
0.4833 | 0.61 | 1000 | 0.8089 | 0.6181 |
0.5733 | 0.91 | 1500 | 0.7617 | 0.5530 |
0.4641 | 1.21 | 2000 | 0.7937 | 0.5731 |
0.4167 | 1.52 | 2500 | 0.7993 | 0.5102 |
0.3713 | 1.82 | 3000 | 0.7541 | 0.5437 |
0.3395 | 2.12 | 3500 | 0.7658 | 0.5148 |
0.2814 | 2.42 | 4000 | 0.7569 | 0.4783 |
0.2698 | 2.73 | 4500 | 0.8126 | 0.5174 |
0.2767 | 3.03 | 5000 | 0.7838 | 0.4676 |
0.2249 | 3.33 | 5500 | 0.8769 | 0.4743 |
0.2452 | 3.64 | 6000 | 0.8586 | 0.4778 |
0.1828 | 3.94 | 6500 | 0.7695 | 0.4528 |
0.1901 | 4.24 | 7000 | 0.7800 | 0.5021 |
0.2062 | 4.55 | 7500 | 0.8107 | 0.4567 |
0.1614 | 4.85 | 8000 | 0.7941 | 0.4094 |
0.1327 | 5.15 | 8500 | 0.7900 | 0.4241 |
0.1405 | 5.45 | 9000 | 0.8017 | 0.3992 |
0.1219 | 5.76 | 9500 | 0.8099 | 0.4043 |
0.1406 | 6.06 | 10000 | 0.8731 | 0.3913 |
0.0806 | 6.36 | 10500 | 0.8387 | 0.3868 |
0.1039 | 6.67 | 11000 | 0.8105 | 0.3905 |
0.0967 | 6.97 | 11500 | 0.7291 | 0.3728 |
0.0846 | 7.27 | 12000 | 0.8128 | 0.4201 |
0.0722 | 7.58 | 12500 | 0.8204 | 0.3751 |
0.0785 | 7.88 | 13000 | 0.7692 | 0.3760 |
0.0647 | 8.18 | 13500 | 0.8294 | 0.3752 |
0.0523 | 8.48 | 14000 | 0.7646 | 0.3763 |
0.0623 | 8.79 | 14500 | 0.7773 | 0.3572 |
0.0477 | 9.09 | 15000 | 0.7379 | 0.3635 |
0.064 | 9.39 | 15500 | 0.7544 | 0.3538 |
0.0321 | 9.7 | 16000 | 0.8118 | 0.3557 |
0.0541 | 10.0 | 16500 | 0.7785 | 0.3534 |
Framework versions
- Transformers 4.26.0
- Pytorch 1.10.0
- Datasets 2.9.0
- Tokenizers 0.13.2