super_large_finetune_M01
This model is a fine-tuned version of facebook/wav2vec2-base on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 2.9906
- Wer: 1.0
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: 20
- 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: 35440
- num_epochs: 200
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 10.0626 | 20.0 | 70880 | 3.0307 | 1.0 |
| 2.5319 | 40.0 | 141760 | 3.0316 | 1.0 |
| 2.4978 | 60.0 | 212640 | 3.0123 | 1.0 |
| 2.4849 | 80.0 | 283520 | 2.9923 | 1.0 |
| 2.4776 | 100.0 | 354400 | 3.0092 | 1.0 |
| 2.4733 | 120.0 | 425280 | 2.9964 | 1.0 |
| 2.4702 | 140.0 | 496160 | 2.9968 | 1.0 |
| 2.4686 | 160.0 | 567040 | 2.9937 | 1.0 |
| 2.4669 | 180.0 | 637920 | 2.9908 | 1.0 |
| 2.4661 | 200.0 | 708800 | 2.9906 | 1.0 |
Framework versions
- Transformers 4.18.0
- Pytorch 1.10.2+cu102
- Datasets 2.3.2
- Tokenizers 0.12.1