<!-- 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-speech-project-subsetData
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: 0.1157
- Wer: 0.2453
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 |
---|---|---|---|---|
14.9506 | 1.67 | 50 | 16.5714 | 1.0 |
7.9883 | 3.33 | 100 | 4.1299 | 1.0 |
3.4027 | 5.0 | 150 | 3.2576 | 1.0 |
3.1425 | 6.67 | 200 | 3.0644 | 1.0 |
3.0198 | 8.33 | 250 | 2.9955 | 1.0 |
2.974 | 10.0 | 300 | 2.9685 | 1.0 |
2.9492 | 11.67 | 350 | 2.9520 | 1.0 |
2.9225 | 13.33 | 400 | 3.0013 | 1.0 |
2.8383 | 15.0 | 450 | 2.7433 | 1.0 |
2.5389 | 16.67 | 500 | 2.1815 | 1.0 |
1.8402 | 18.33 | 550 | 1.2757 | 0.9653 |
0.926 | 20.0 | 600 | 0.4806 | 0.4453 |
0.4593 | 21.67 | 650 | 0.3497 | 0.376 |
0.3047 | 23.33 | 700 | 0.1765 | 0.2507 |
0.2407 | 25.0 | 750 | 0.1152 | 0.216 |
0.1867 | 26.67 | 800 | 0.2214 | 0.2987 |
0.1637 | 28.33 | 850 | 0.1050 | 0.2453 |
0.1599 | 30.0 | 900 | 0.1157 | 0.2453 |
Framework versions
- Transformers 4.28.1
- Pytorch 2.0.0+cu118
- Datasets 1.18.3
- Tokenizers 0.13.3