CeLLaTe-ner-2class-reinit-llrd-pubmedbert-tapt-tokenizer-adapted_wr-0.06
This model is a fine-tuned version of Mardiyyah/CeLLaTe-tapt-pubmedbert-tokenizer-adapted_v2_lr-3e5 on the OTAR3088/CeLLaTe_V3.2_contracted_ent_IOB dataset. It achieves the following results on the evaluation set:
- Loss: 0.0727
- Precision: 0.7829
- Recall: 0.7168
- F1: 0.7484
- 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: 3e-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: 0.06
- num_epochs: 20
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Precision | Recall | F1 | Accuracy |
|---|---|---|---|---|---|---|---|
| 0.3696 | 1.0 | 263 | 0.0713 | 0.4648 | 0.5377 | 0.4986 | 0.9744 |
| 0.0449 | 2.0 | 526 | 0.0586 | 0.7649 | 0.6838 | 0.7221 | 0.9829 |
| 0.026 | 3.0 | 789 | 0.0557 | 0.7508 | 0.7234 | 0.7368 | 0.9828 |
| 0.017 | 4.0 | 1052 | 0.0665 | 0.7527 | 0.7144 | 0.7330 | 0.9832 |
| 0.012 | 5.0 | 1315 | 0.0724 | 0.7829 | 0.7168 | 0.7484 | 0.9837 |
| 0.0093 | 6.0 | 1578 | 0.0822 | 0.7435 | 0.6856 | 0.7134 | 0.9816 |
| 0.0068 | 7.0 | 1841 | 0.0859 | 0.7765 | 0.7054 | 0.7393 | 0.9833 |
| 0.005 | 8.0 | 2104 | 0.0925 | 0.7959 | 0.7054 | 0.7479 | 0.9831 |
Framework versions
- Transformers 4.48.2
- Pytorch 2.4.1+cu121
- Datasets 3.0.2
- Tokenizers 0.21.0