CeLLaTe-pubmedbert-tapt-base

This model is a fine-tuned version of microsoft/BiomedNLP-BiomedBERT-base-uncased-abstract-fulltext on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.9900
• Accuracy: 0.7793
• Perplexity: 2.6912

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: 50
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Perplexity
No log	1.0	14	1.0006	0.7773	2.7199
No log	2.0	28	0.9782	0.7805	2.6596
No log	3.0	42	0.9766	0.7810	2.6553
No log	4.0	56	0.9922	0.7769	2.6972
No log	5.0	70	0.9743	0.7839	2.6494
No log	6.0	84	0.9940	0.7780	2.7019
No log	7.0	98	0.9677	0.7798	2.6318
1.0665	8.0	112	0.9465	0.7851	2.5766
1.0665	9.0	126	1.0023	0.7738	2.7245
1.0665	10.0	140	0.9817	0.7796	2.6690
1.0665	11.0	154	0.9689	0.7815	2.6351

Framework versions

• Transformers 4.48.2
• Pytorch 2.4.1+cu121
• Datasets 3.0.2
• Tokenizers 0.21.0