CeLLaTe-ner-3class-bioformer16l-tapt-tokenizer-original-baseline
This model is a fine-tuned version of Mardiyyah/CeLLaTe-tapt-bioformer16l-tokenizer-original-baseline on the OTAR3088/CeLLaTe-ner-3class-iob_final dataset. It achieves the following results on the evaluation set:
- Loss: 0.1011
- Precision: 0.7839
- Recall: 0.7558
- Micro F1: 0.7696
- Weighted F1: 0.7697
- Macro F1: 0.7605
- Accuracy: 0.9810
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 |
|---|---|---|---|---|---|---|---|---|---|
| 0.4197 | 1.0 | 263 | 0.1065 | 0.3777 | 0.4234 | 0.3993 | 0.4005 | 0.3666 | 0.9628 |
| 0.0812 | 2.0 | 526 | 0.0739 | 0.7431 | 0.7007 | 0.7213 | 0.7203 | 0.7082 | 0.9787 |
| 0.0489 | 3.0 | 789 | 0.0731 | 0.7443 | 0.6936 | 0.7180 | 0.7181 | 0.7111 | 0.9786 |
| 0.0353 | 4.0 | 1052 | 0.0816 | 0.7927 | 0.7125 | 0.7505 | 0.7475 | 0.7342 | 0.9799 |
| 0.0275 | 5.0 | 1315 | 0.0807 | 0.7766 | 0.7271 | 0.7510 | 0.7506 | 0.7411 | 0.9800 |
| 0.0219 | 6.0 | 1578 | 0.0921 | 0.7278 | 0.6920 | 0.7095 | 0.7107 | 0.7041 | 0.9777 |
| 0.0181 | 7.0 | 1841 | 0.0907 | 0.7282 | 0.7176 | 0.7229 | 0.7244 | 0.7171 | 0.9780 |
| 0.0153 | 8.0 | 2104 | 0.1052 | 0.7320 | 0.6810 | 0.7056 | 0.7066 | 0.7006 | 0.9775 |
| 0.0123 | 9.0 | 2367 | 0.0961 | 0.7497 | 0.7397 | 0.7446 | 0.7452 | 0.7387 | 0.9791 |
| 0.0113 | 10.0 | 2630 | 0.0939 | 0.7597 | 0.7448 | 0.7522 | 0.7527 | 0.7454 | 0.9796 |
| 0.0098 | 11.0 | 2893 | 0.0974 | 0.7445 | 0.7483 | 0.7464 | 0.7482 | 0.7396 | 0.9795 |
| 0.0085 | 12.0 | 3156 | 0.1007 | 0.7839 | 0.7558 | 0.7696 | 0.7697 | 0.7605 | 0.9810 |
| 0.0078 | 13.0 | 3419 | 0.1090 | 0.7228 | 0.7314 | 0.7271 | 0.7291 | 0.7224 | 0.9782 |
| 0.0073 | 14.0 | 3682 | 0.1059 | 0.7746 | 0.7416 | 0.7577 | 0.7589 | 0.7514 | 0.9804 |
| 0.0062 | 15.0 | 3945 | 0.1060 | 0.7713 | 0.7625 | 0.7669 | 0.7679 | 0.7581 | 0.9806 |
| 0.0058 | 16.0 | 4208 | 0.1092 | 0.7483 | 0.7507 | 0.7495 | 0.7512 | 0.7442 | 0.9796 |
| 0.0052 | 17.0 | 4471 | 0.1136 | 0.7791 | 0.7460 | 0.7622 | 0.7626 | 0.7540 | 0.9802 |
Framework versions
- Transformers 4.48.2
- Pytorch 2.4.1+cu121
- Datasets 3.0.2
- Tokenizers 0.21.0