<!-- 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. -->
sd5
This model is a fine-tuned version of Theju/sd5 on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.0472
- Wer: 1.1713
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: 4
- 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: 7
- mixed_precision_training: Native AMP
Training results
Training Loss | Epoch | Step | Validation Loss | Wer |
---|---|---|---|---|
2.536 | 0.68 | 500 | 2.4893 | 2.5109 |
2.0499 | 1.36 | 1000 | 1.3903 | 2.3178 |
1.2147 | 2.04 | 1500 | 0.5195 | 1.8536 |
0.6346 | 2.72 | 2000 | 0.1633 | 1.2617 |
0.3675 | 3.41 | 2500 | 0.1510 | 1.3115 |
0.2561 | 4.09 | 3000 | 0.1246 | 1.6760 |
0.1612 | 4.77 | 3500 | 0.0781 | 1.4330 |
0.111 | 5.45 | 4000 | 0.0811 | 1.3676 |
0.0669 | 6.13 | 4500 | 0.0582 | 1.1900 |
0.0575 | 6.81 | 5000 | 0.0472 | 1.1713 |
Framework versions
- Transformers 4.17.0
- Pytorch 1.13.1+cu116
- Datasets 1.18.3
- Tokenizers 0.13.2