ctrlv-wav2vec2-tokenizer
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: 0.3967
- Wer: 0.3138
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: 32
- 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 |
|---|---|---|---|---|
| 3.4359 | 3.45 | 500 | 1.3595 | 0.9159 |
| 0.5692 | 6.9 | 1000 | 0.4332 | 0.4036 |
| 0.2198 | 10.34 | 1500 | 0.4074 | 0.3678 |
| 0.1314 | 13.79 | 2000 | 0.3480 | 0.3409 |
| 0.0929 | 17.24 | 2500 | 0.3714 | 0.3346 |
| 0.0692 | 20.69 | 3000 | 0.3977 | 0.3224 |
| 0.0542 | 24.14 | 3500 | 0.4068 | 0.3187 |
| 0.0422 | 27.59 | 4000 | 0.3967 | 0.3138 |
Framework versions
- Transformers 4.11.3
- Pytorch 1.10.0+cu111
- Datasets 1.18.3
- Tokenizers 0.10.3