CeLLaTe-tapt-pubmedbert-tokenizer-adapted_v2_lr-5e5
This model is a fine-tuned version of microsoft/BiomedNLP-BiomedBERT-base-uncased-abstract-fulltext on the Mardiyyah/TAPT_CeLLaTe2.0 dataset. It achieves the following results on the evaluation set:
- Loss: 1.1756
- Accuracy: 0.7630
- Perplexity: 3.2401
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: 5e-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 |
|---|---|---|---|---|---|
| 1.3568 | 1.0 | 14 | 1.1951 | 0.7597 | 3.3038 |
| 1.3615 | 2.0 | 28 | 1.1865 | 0.7620 | 3.2756 |
| 1.3234 | 3.0 | 42 | 1.1876 | 0.7577 | 3.2792 |
| 1.3074 | 4.0 | 56 | 1.2069 | 0.7564 | 3.3432 |
| 1.2786 | 5.0 | 70 | 1.2031 | 0.7565 | 3.3303 |
| 1.261 | 6.0 | 84 | 1.1954 | 0.7546 | 3.3048 |
| 1.2397 | 7.0 | 98 | 1.1487 | 0.7636 | 3.1540 |
| 1.2315 | 8.0 | 112 | 1.2123 | 0.7520 | 3.3613 |
| 1.1975 | 9.0 | 126 | 1.1515 | 0.7574 | 3.1630 |
| 1.1853 | 10.0 | 140 | 1.1865 | 0.7563 | 3.2756 |
| 1.1912 | 11.0 | 154 | 1.1648 | 0.7541 | 3.2052 |
| 1.1469 | 12.0 | 168 | 1.1832 | 0.7563 | 3.2648 |
Framework versions
- Transformers 4.48.2
- Pytorch 2.4.1+cu121
- Datasets 3.0.2
- Tokenizers 0.21.0