shan-stt-v3
This model is a fine-tuned version of k1ngtai/shan-stt-v2 on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.2857
- Model Preparation Time: 0.0053
- Wer: 89.5425
- Cer: 19.6965
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: 1e-05
- train_batch_size: 16
- eval_batch_size: 8
- seed: 42
- optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 100
- training_steps: 2000
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Model Preparation Time | Wer | Cer |
|---|---|---|---|---|---|---|
| 0.0042 | 12.5 | 250 | 0.2420 | 0.0053 | 92.1569 | 23.2369 |
| 0.0002 | 25.0 | 500 | 0.2624 | 0.0053 | 88.8889 | 20.4552 |
| 0.0000 | 37.5 | 750 | 0.2754 | 0.0053 | 90.1961 | 20.0056 |
| 0.0000 | 50.0 | 1000 | 0.2817 | 0.0053 | 90.1961 | 19.8932 |
| 0.0000 | 62.5 | 1250 | 0.2857 | 0.0053 | 89.5425 | 19.6965 |
| 0.0000 | 75.0 | 1500 | 0.2886 | 0.0053 | 89.5425 | 19.9494 |
| 0.0000 | 87.5 | 1750 | 0.2907 | 0.0053 | 89.5425 | 20.0899 |
| 0.0000 | 100.0 | 2000 | 0.2912 | 0.0053 | 90.1961 | 20.0056 |
Framework versions
- Transformers 5.0.0
- Pytorch 2.10.0+cu128
- Datasets 3.6.0
- Tokenizers 0.22.2