CeLLaTe-ner-2class-reinit_llrd-pubmedbert-tapt-tokenizer-adapted-spanmask-gazetters-lr_4.169
This model is a fine-tuned version of Mardiyyah/CeLLaTe-tapt-pubmedbert-tokenizer-adapted-spanmask on the OTAR3088/CeLLaTe-ner-2class-iob_final dataset. It achieves the following results on the evaluation set:
- Loss: 0.1085
- Precision: 0.8014
- Recall: 0.7787
- Micro F1: 0.7899
- Weighted F1: 0.7900
- Macro F1: 0.7958
- Accuracy: 0.9855
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: 4.1695990556e-05
- train_batch_size: 16
- eval_batch_size: 16
- seed: 2026
- gradient_accumulation_steps: 2
- total_train_batch_size: 32
- 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: cosine
- lr_scheduler_warmup_ratio: 0.03
- num_epochs: 20
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Precision | Recall | Micro F1 | Weighted F1 | Macro F1 | Accuracy |
|---|---|---|---|---|---|---|---|---|---|
| 0.156 | 1.0 | 1052 | 0.0645 | 0.7314 | 0.7468 | 0.7391 | 0.7399 | 0.7520 | 0.9832 |
| 0.0137 | 2.0 | 2104 | 0.0780 | 0.8041 | 0.7132 | 0.7559 | 0.7559 | 0.7666 | 0.9838 |
| 0.006 | 3.0 | 3156 | 0.0818 | 0.7424 | 0.7679 | 0.7550 | 0.7553 | 0.7650 | 0.9826 |
| 0.0038 | 4.0 | 4208 | 0.0947 | 0.7485 | 0.7697 | 0.7590 | 0.7592 | 0.7684 | 0.9830 |
| 0.0027 | 5.0 | 5260 | 0.1075 | 0.7372 | 0.7775 | 0.7568 | 0.7573 | 0.7670 | 0.9826 |
| 0.0021 | 6.0 | 6312 | 0.1112 | 0.7202 | 0.7956 | 0.756 | 0.7564 | 0.7630 | 0.9823 |
| 0.0019 | 7.0 | 7364 | 0.1120 | 0.7606 | 0.7354 | 0.7478 | 0.7478 | 0.7513 | 0.9828 |
| 0.0013 | 8.0 | 8416 | 0.1093 | 0.8014 | 0.7787 | 0.7899 | 0.7900 | 0.7958 | 0.9855 |
| 0.0015 | 9.0 | 9468 | 0.1177 | 0.8046 | 0.7577 | 0.7804 | 0.7805 | 0.7865 | 0.9848 |
| 0.0009 | 10.0 | 10520 | 0.1174 | 0.7576 | 0.7781 | 0.7677 | 0.7679 | 0.7729 | 0.9835 |
| 0.0008 | 11.0 | 11572 | 0.1164 | 0.7782 | 0.7745 | 0.7764 | 0.7768 | 0.7843 | 0.9844 |
| 0.0007 | 12.0 | 12624 | 0.1481 | 0.7200 | 0.7962 | 0.7561 | 0.7570 | 0.7660 | 0.9819 |
| 0.0006 | 13.0 | 13676 | 0.1353 | 0.7865 | 0.7865 | 0.7865 | 0.7866 | 0.7906 | 0.9843 |
Framework versions
- Transformers 4.48.2
- Pytorch 2.4.1+cu121
- Datasets 3.0.2
- Tokenizers 0.21.0