CeLLaTe-tapt-pubmedbert-tokenizer-adapted_v2_lr-3e5
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: 1.1493
- Accuracy: 0.7653
- Perplexity: 3.1561
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 |
|---|---|---|---|---|---|
| 1.3584 | 1.0 | 14 | 1.2016 | 0.7595 | 3.3253 |
| 1.3699 | 2.0 | 28 | 1.1943 | 0.7615 | 3.3012 |
| 1.3374 | 3.0 | 42 | 1.1951 | 0.7587 | 3.3038 |
| 1.3226 | 4.0 | 56 | 1.2118 | 0.7565 | 3.3596 |
| 1.2976 | 5.0 | 70 | 1.2074 | 0.7574 | 3.3449 |
| 1.283 | 6.0 | 84 | 1.1959 | 0.7552 | 3.3066 |
| 1.2645 | 7.0 | 98 | 1.1472 | 0.7638 | 3.1494 |
| 1.2597 | 8.0 | 112 | 1.2038 | 0.7514 | 3.3326 |
| 1.2291 | 9.0 | 126 | 1.1464 | 0.7597 | 3.1469 |
| 1.2186 | 10.0 | 140 | 1.1744 | 0.7571 | 3.2362 |
| 1.2329 | 11.0 | 154 | 1.1545 | 0.7565 | 3.1724 |
| 1.1898 | 12.0 | 168 | 1.1718 | 0.7589 | 3.2278 |
| 1.1865 | 13.0 | 182 | 1.2333 | 0.7520 | 3.4325 |
| 1.1772 | 14.0 | 196 | 1.1856 | 0.7567 | 3.2727 |
Framework versions
- Transformers 4.48.2
- Pytorch 2.4.1+cu121
- Datasets 3.0.2
- Tokenizers 0.21.0