medical_bert_tiny_asymmetric_old_dataset_v1
This model is a fine-tuned version of prajjwal1/bert-tiny on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.0130
- Accuracy: 0.9885
- Recall Weighted: 0.9885
- Precision Weighted: 0.9885
- F1 Weighted: 0.9885
- F1 Macro: 0.9834
- F1 Prescription: 0.9730
- F1 Lab Report: 0.9771
- F1 Others: 1.0
- False Alarm Rate: 0.0
- Cross Class Error: 0.0248
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: 0.0002
- train_batch_size: 8
- eval_batch_size: 64
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 16
- optimizer: Use OptimizerNames.ADAMW_TORCH 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: 0.1
- num_epochs: 5
- label_smoothing_factor: 0.1
Training results
| Training Loss | Epoch | Step | Validation Loss | Accuracy | Recall Weighted | Precision Weighted | F1 Weighted | F1 Macro | F1 Prescription | F1 Lab Report | F1 Others | False Alarm Rate | Cross Class Error |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.0885 | 1.0 | 892 | 0.0208 | 0.9837 | 0.9837 | 0.9840 | 0.9837 | 0.9764 | 0.9611 | 0.9680 | 1.0 | 0.0 | 0.0351 |
| 0.1100 | 2.0 | 1784 | 0.0176 | 0.9847 | 0.9847 | 0.9849 | 0.9846 | 0.9778 | 0.9635 | 0.9698 | 1.0 | 0.0 | 0.0331 |
| 0.0162 | 3.0 | 2676 | 0.0169 | 0.9851 | 0.9851 | 0.9857 | 0.9851 | 0.9784 | 0.9643 | 0.9709 | 1.0 | 0.0 | 0.0320 |
| 0.0538 | 4.0 | 3568 | 0.0131 | 0.9880 | 0.9880 | 0.9881 | 0.9880 | 0.9827 | 0.9718 | 0.9762 | 1.0 | 0.0 | 0.0258 |
| 0.0004 | 5.0 | 4460 | 0.0130 | 0.9885 | 0.9885 | 0.9885 | 0.9885 | 0.9834 | 0.9730 | 0.9771 | 1.0 | 0.0 | 0.0248 |
Framework versions
- PEFT 0.18.1
- Transformers 5.0.0
- Pytorch 2.10.0+cu128
- Datasets 4.8.5
- Tokenizers 0.22.2