wavlm-libri-clean-100h-base
This model is a fine-tuned version of microsoft/wavlm-base on the LIBRISPEECH_ASR - CLEAN dataset. It achieves the following results on the evaluation set:
- Loss: 0.0955
- Wer: 0.0773
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: 0.0003
- train_batch_size: 16
- eval_batch_size: 16
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- num_epochs: 1.0
- mixed_precision_training: Native AMP
Training results
Training Loss | Epoch | Step | Validation Loss | Wer |
---|---|---|---|---|
2.8664 | 0.17 | 300 | 2.8439 | 1.0 |
0.5009 | 0.34 | 600 | 0.2709 | 0.2162 |
0.2056 | 0.5 | 900 | 0.1934 | 0.1602 |
0.1648 | 0.67 | 1200 | 0.1576 | 0.1306 |
0.1922 | 0.84 | 1500 | 0.1358 | 0.1114 |
0.093 | 1.01 | 1800 | 0.1277 | 0.1035 |
0.0652 | 1.18 | 2100 | 0.1251 | 0.1005 |
0.0848 | 1.35 | 2400 | 0.1188 | 0.0964 |
0.0706 | 1.51 | 2700 | 0.1091 | 0.0905 |
0.0846 | 1.68 | 3000 | 0.1018 | 0.0840 |
0.0684 | 1.85 | 3300 | 0.0978 | 0.0809 |
Framework versions
- Transformers 4.15.0
- Pytorch 1.9.1
- Datasets 1.18.0
- Tokenizers 0.10.3
- Downloads last month
- 24
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social
visibility and check back later, or deploy to Inference Endpoints (dedicated)
instead.