CeLLaTe-ner-3class-pubmedbert-tapt-tokenizer-adapted-spanmask
This model is a fine-tuned version of Mardiyyah/CeLLaTe-tapt-pubmedbert-tokenizer-adapted-spanmask on the OTAR3088/CeLLaTe-ner-3class-iob_final dataset. It achieves the following results on the evaluation set:
- Loss: 0.0801
- Precision: 0.7559
- Recall: 0.7419
- Micro F1: 0.7488
- Weighted F1: 0.7488
- Macro F1: 0.7453
- Accuracy: 0.9823
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 |
|---|---|---|---|---|---|---|---|---|---|
| 0.3793 | 1.0 | 263 | 0.0956 | 0.3836 | 0.5290 | 0.4447 | 0.4350 | 0.4047 | 0.9693 |
| 0.0599 | 2.0 | 526 | 0.0685 | 0.7314 | 0.6819 | 0.7058 | 0.7066 | 0.7070 | 0.9798 |
| 0.0349 | 3.0 | 789 | 0.0735 | 0.7028 | 0.7193 | 0.7110 | 0.7109 | 0.7115 | 0.9796 |
| 0.0246 | 4.0 | 1052 | 0.0803 | 0.7559 | 0.7419 | 0.7488 | 0.7488 | 0.7453 | 0.9823 |
| 0.0189 | 5.0 | 1315 | 0.0852 | 0.7606 | 0.7060 | 0.7323 | 0.7317 | 0.7285 | 0.9810 |
| 0.0146 | 6.0 | 1578 | 0.0929 | 0.7200 | 0.6875 | 0.7034 | 0.7040 | 0.6986 | 0.9796 |
| 0.0114 | 7.0 | 1841 | 0.0963 | 0.7157 | 0.7286 | 0.7221 | 0.7221 | 0.7224 | 0.9802 |
| 0.0089 | 8.0 | 2104 | 0.0995 | 0.7563 | 0.7291 | 0.7424 | 0.7415 | 0.7400 | 0.9817 |
| 0.0075 | 9.0 | 2367 | 0.0999 | 0.6893 | 0.7501 | 0.7184 | 0.7200 | 0.7201 | 0.9796 |
Framework versions
- Transformers 4.48.2
- Pytorch 2.4.1+cu121
- Datasets 3.0.2
- Tokenizers 0.21.0