CeLLaTe-ner-2class-pubmedbert-tapt-tokenizer-adapted-spanmask-lr_2.37
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: 0.0854
- Precision: 0.7459
- Recall: 0.7643
- Micro F1: 0.7550
- Weighted F1: 0.7554
- Macro F1: 0.7691
- Accuracy: 0.9837
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: 2.3706685643807145e-05
- train_batch_size: 32
- eval_batch_size: 16
- seed: 10472999
- gradient_accumulation_steps: 2
- total_train_batch_size: 64
- 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: 0.05
- 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 |
|---|---|---|---|---|---|---|---|---|---|
| 0.6564 | 1.0 | 132 | 0.1156 | 0.2252 | 0.2038 | 0.2140 | 0.1787 | 0.1265 | 0.9545 |
| 0.0783 | 2.0 | 264 | 0.0548 | 0.7026 | 0.7017 | 0.7022 | 0.7028 | 0.7232 | 0.9820 |
| 0.0362 | 3.0 | 396 | 0.0590 | 0.7170 | 0.6933 | 0.7050 | 0.7056 | 0.7227 | 0.9820 |
| 0.0258 | 4.0 | 528 | 0.0644 | 0.7311 | 0.7162 | 0.7236 | 0.7242 | 0.7366 | 0.9827 |
| 0.0191 | 5.0 | 660 | 0.0626 | 0.7341 | 0.7270 | 0.7305 | 0.7308 | 0.7402 | 0.9834 |
| 0.0152 | 6.0 | 792 | 0.0681 | 0.7593 | 0.7150 | 0.7365 | 0.7367 | 0.7458 | 0.9831 |
| 0.0118 | 7.0 | 924 | 0.0709 | 0.7411 | 0.7486 | 0.7448 | 0.7451 | 0.7533 | 0.9835 |
| 0.0092 | 8.0 | 1056 | 0.0790 | 0.7654 | 0.7162 | 0.7400 | 0.7403 | 0.7518 | 0.9833 |
| 0.0074 | 9.0 | 1188 | 0.0743 | 0.7520 | 0.7384 | 0.7451 | 0.7452 | 0.7536 | 0.9840 |
| 0.0062 | 10.0 | 1320 | 0.0801 | 0.7458 | 0.7252 | 0.7354 | 0.7360 | 0.7501 | 0.9834 |
| 0.0053 | 11.0 | 1452 | 0.0843 | 0.7564 | 0.7450 | 0.7507 | 0.7509 | 0.7617 | 0.9842 |
| 0.0045 | 12.0 | 1584 | 0.0844 | 0.7452 | 0.7511 | 0.7481 | 0.7488 | 0.7618 | 0.9835 |
| 0.0038 | 13.0 | 1716 | 0.0859 | 0.7459 | 0.7643 | 0.7550 | 0.7554 | 0.7691 | 0.9837 |
| 0.0037 | 14.0 | 1848 | 0.0914 | 0.7714 | 0.7306 | 0.7505 | 0.7506 | 0.7621 | 0.9837 |
| 0.0032 | 15.0 | 1980 | 0.0891 | 0.7531 | 0.7354 | 0.7441 | 0.7448 | 0.7594 | 0.9837 |
| 0.003 | 16.0 | 2112 | 0.0925 | 0.7601 | 0.7336 | 0.7466 | 0.7470 | 0.7584 | 0.9837 |
| 0.0028 | 17.0 | 2244 | 0.0917 | 0.7465 | 0.7420 | 0.7443 | 0.7447 | 0.7567 | 0.9839 |
| 0.0028 | 18.0 | 2376 | 0.0928 | 0.7560 | 0.7378 | 0.7468 | 0.7472 | 0.7596 | 0.9837 |
Framework versions
- Transformers 4.48.2
- Pytorch 2.4.1+cu121
- Datasets 3.0.2
- Tokenizers 0.21.0