CeLLaTe-ner-3class-bioformer16l-tapt-tokenizer-original-wwmask
This model is a fine-tuned version of Mardiyyah/CeLLaTe-tapt-bioformer16l-tokenizer-original-wwmask on the OTAR3088/CeLLaTe-ner-3class-iob_final dataset. It achieves the following results on the evaluation set:
- Loss: 0.1025
- Precision: 0.7686
- Recall: 0.7562
- Micro F1: 0.7624
- Weighted F1: 0.7623
- Macro F1: 0.7515
- Accuracy: 0.9804
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 |
|---|---|---|---|---|---|---|---|---|---|
| 0.418 | 1.0 | 263 | 0.1076 | 0.3909 | 0.4415 | 0.4146 | 0.4128 | 0.3775 | 0.9628 |
| 0.0808 | 2.0 | 526 | 0.0762 | 0.7292 | 0.6597 | 0.6927 | 0.6919 | 0.6815 | 0.9775 |
| 0.049 | 3.0 | 789 | 0.0718 | 0.7163 | 0.6959 | 0.7060 | 0.7064 | 0.7005 | 0.9780 |
| 0.036 | 4.0 | 1052 | 0.0821 | 0.7710 | 0.6774 | 0.7212 | 0.7180 | 0.7053 | 0.9784 |
| 0.0275 | 5.0 | 1315 | 0.0797 | 0.7620 | 0.7176 | 0.7391 | 0.7387 | 0.7305 | 0.9796 |
| 0.0217 | 6.0 | 1578 | 0.0882 | 0.7241 | 0.7089 | 0.7164 | 0.7172 | 0.7120 | 0.9779 |
| 0.0185 | 7.0 | 1841 | 0.0876 | 0.7423 | 0.7271 | 0.7346 | 0.7350 | 0.7274 | 0.9787 |
| 0.0149 | 8.0 | 2104 | 0.0989 | 0.7569 | 0.7089 | 0.7322 | 0.7316 | 0.7232 | 0.9789 |
| 0.0126 | 9.0 | 2367 | 0.0917 | 0.7483 | 0.7519 | 0.7501 | 0.7501 | 0.7396 | 0.9794 |
| 0.0115 | 10.0 | 2630 | 0.0952 | 0.7316 | 0.7184 | 0.7250 | 0.7255 | 0.7182 | 0.9787 |
| 0.0098 | 11.0 | 2893 | 0.0999 | 0.7415 | 0.7298 | 0.7356 | 0.7354 | 0.7270 | 0.9789 |
| 0.0086 | 12.0 | 3156 | 0.0973 | 0.7631 | 0.7550 | 0.7591 | 0.7590 | 0.7500 | 0.9804 |
| 0.0079 | 13.0 | 3419 | 0.1056 | 0.7129 | 0.7294 | 0.7210 | 0.7217 | 0.7138 | 0.9777 |
| 0.0074 | 14.0 | 3682 | 0.1067 | 0.7753 | 0.7432 | 0.7589 | 0.7588 | 0.7483 | 0.9802 |
| 0.0065 | 15.0 | 3945 | 0.1025 | 0.7686 | 0.7562 | 0.7624 | 0.7623 | 0.7515 | 0.9804 |
| 0.0059 | 16.0 | 4208 | 0.1073 | 0.7573 | 0.7511 | 0.7542 | 0.7543 | 0.7470 | 0.9799 |
| 0.0054 | 17.0 | 4471 | 0.1105 | 0.7607 | 0.7436 | 0.7520 | 0.7523 | 0.7438 | 0.9801 |
| 0.0052 | 18.0 | 4734 | 0.1106 | 0.7472 | 0.7487 | 0.7480 | 0.7486 | 0.7402 | 0.9796 |
| 0.0049 | 19.0 | 4997 | 0.1130 | 0.7488 | 0.7456 | 0.7472 | 0.7473 | 0.7392 | 0.9796 |
| 0.0047 | 20.0 | 5260 | 0.1123 | 0.7542 | 0.7519 | 0.7531 | 0.7531 | 0.7448 | 0.9799 |
Framework versions
- Transformers 4.48.2
- Pytorch 2.4.1+cu121
- Datasets 3.0.2
- Tokenizers 0.21.0