CeLLaTe-ner-2class-tapt-pubmedbert-tokenizer-adapted-wwmask
This model is a fine-tuned version of Mardiyyah/CeLLaTe-tapt-pubmedbert-tokenizer-adapted-wwmask on the OTAR3088/CeLLaTe-ner-2class-iob_final dataset. It achieves the following results on the evaluation set:
- Loss: 0.1032
- Precision: 0.7757
- Recall: 0.7571
- Micro F1: 0.7663
- Weighted F1: 0.7667
- Macro F1: 0.7765
- Accuracy: 0.9842
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: 2e-05
- train_batch_size: 16
- eval_batch_size: 16
- seed: 3407
- 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: linear
- lr_scheduler_warmup_ratio: 0.01
- 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 |
|---|---|---|---|---|---|---|---|---|---|
| 0.2519 | 1.0 | 263 | 0.0602 | 0.6575 | 0.6927 | 0.6747 | 0.6752 | 0.6828 | 0.9817 |
| 0.041 | 2.0 | 526 | 0.0563 | 0.7232 | 0.7150 | 0.7191 | 0.7203 | 0.7373 | 0.9823 |
| 0.0265 | 3.0 | 789 | 0.0593 | 0.7150 | 0.7348 | 0.7248 | 0.7258 | 0.7434 | 0.9826 |
| 0.0195 | 4.0 | 1052 | 0.0623 | 0.7512 | 0.7318 | 0.7414 | 0.7415 | 0.7493 | 0.9832 |
| 0.0147 | 5.0 | 1315 | 0.0719 | 0.7758 | 0.6867 | 0.7285 | 0.7285 | 0.7360 | 0.9823 |
| 0.0121 | 6.0 | 1578 | 0.0624 | 0.7512 | 0.7462 | 0.7487 | 0.7489 | 0.7541 | 0.9841 |
| 0.009 | 7.0 | 1841 | 0.0771 | 0.7665 | 0.7402 | 0.7531 | 0.7533 | 0.7609 | 0.9835 |
| 0.0067 | 8.0 | 2104 | 0.0754 | 0.7677 | 0.7372 | 0.7521 | 0.7521 | 0.7567 | 0.9834 |
| 0.006 | 9.0 | 2367 | 0.0819 | 0.7209 | 0.7673 | 0.7434 | 0.7441 | 0.7558 | 0.9830 |
| 0.0047 | 10.0 | 2630 | 0.0844 | 0.7631 | 0.7420 | 0.7524 | 0.7529 | 0.7649 | 0.9838 |
| 0.0037 | 11.0 | 2893 | 0.0870 | 0.7586 | 0.7595 | 0.7590 | 0.7596 | 0.7697 | 0.9837 |
| 0.0036 | 12.0 | 3156 | 0.0877 | 0.7702 | 0.7498 | 0.7599 | 0.7604 | 0.7713 | 0.9841 |
| 0.0028 | 13.0 | 3419 | 0.1001 | 0.7598 | 0.7589 | 0.7593 | 0.7595 | 0.7693 | 0.9833 |
| 0.0025 | 14.0 | 3682 | 0.0959 | 0.7585 | 0.7517 | 0.7551 | 0.7556 | 0.7659 | 0.9839 |
| 0.0024 | 15.0 | 3945 | 0.0981 | 0.7618 | 0.7541 | 0.7579 | 0.7586 | 0.7712 | 0.9839 |
| 0.0019 | 16.0 | 4208 | 0.0988 | 0.7592 | 0.7565 | 0.7578 | 0.7584 | 0.7686 | 0.9837 |
| 0.0019 | 17.0 | 4471 | 0.1003 | 0.7854 | 0.7462 | 0.7653 | 0.7657 | 0.7763 | 0.9844 |
| 0.0017 | 18.0 | 4734 | 0.1010 | 0.7725 | 0.7595 | 0.7659 | 0.7662 | 0.7748 | 0.9840 |
| 0.0015 | 19.0 | 4997 | 0.1042 | 0.7792 | 0.7535 | 0.7661 | 0.7665 | 0.7763 | 0.9842 |
| 0.0016 | 20.0 | 5260 | 0.1037 | 0.7762 | 0.7571 | 0.7665 | 0.7669 | 0.7767 | 0.9842 |
Framework versions
- Transformers 4.48.2
- Pytorch 2.4.1+cu121
- Datasets 3.0.2
- Tokenizers 0.21.0