CeLLaTe-tapt-pubmedbert-tokenizer-original-baseline-combinedData
This model is a fine-tuned version of microsoft/BiomedNLP-BiomedBERT-base-uncased-abstract-fulltext on the Mardiyyah/Combined_TAPT_dataset dataset. It achieves the following results on the evaluation set:
- Loss: 0.9904
- Accuracy: 0.7745
- Perplexity: 2.6923
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: 32
- eval_batch_size: 32
- seed: 3407
- gradient_accumulation_steps: 2
- total_train_batch_size: 64
- optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-06 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: linear
- lr_scheduler_warmup_ratio: 0.06
- num_epochs: 100
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Accuracy | Perplexity |
|---|---|---|---|---|---|
| 1.1406 | 1.0 | 28 | 1.0306 | 0.7735 | 2.8028 |
| 1.1214 | 2.0 | 56 | 0.9953 | 0.7771 | 2.7055 |
| 1.1017 | 3.0 | 84 | 0.9777 | 0.7752 | 2.6582 |
| 1.0798 | 4.0 | 112 | 0.9868 | 0.7746 | 2.6826 |
| 1.0642 | 5.0 | 140 | 0.9592 | 0.7816 | 2.6096 |
| 1.0604 | 6.0 | 168 | 0.9888 | 0.7742 | 2.6880 |
| 1.0397 | 7.0 | 196 | 0.9889 | 0.7761 | 2.6884 |
| 1.039 | 8.0 | 224 | 0.9815 | 0.7780 | 2.6683 |
| 1.0295 | 9.0 | 252 | 0.9945 | 0.7729 | 2.7035 |
| 1.0088 | 10.0 | 280 | 0.9534 | 0.7820 | 2.5945 |
| 1.0021 | 11.0 | 308 | 0.9779 | 0.7746 | 2.6590 |
| 0.9926 | 12.0 | 336 | 0.9840 | 0.7751 | 2.6750 |
| 0.9761 | 13.0 | 364 | 0.9873 | 0.7769 | 2.6840 |
| 0.9724 | 14.0 | 392 | 0.9602 | 0.7809 | 2.6121 |
| 0.9665 | 15.0 | 420 | 0.9978 | 0.7726 | 2.7123 |
Framework versions
- Transformers 4.48.2
- Pytorch 2.4.1+cu121
- Datasets 3.0.2
- Tokenizers 0.21.0