CeLLaTe-tapt-pubmedbert-tokenizer-adapted-spanmask-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: 2.2865
• Accuracy: 0.5769
• Perplexity: 9.8401

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
2.9396	1.0	28	2.5275	0.5599	12.5225
2.6425	2.0	56	2.4571	0.5677	11.6704
2.5762	3.0	84	2.3612	0.5702	10.6040
2.5444	4.0	112	2.3182	0.5815	10.1570
2.533	5.0	140	2.3575	0.5703	10.5645
2.4781	6.0	168	2.3388	0.5754	10.3687
2.4447	7.0	196	2.3351	0.5748	10.3308
2.4447	8.0	224	2.3080	0.5757	10.0540
2.4373	9.0	252	2.2718	0.5796	9.6969
2.4251	10.0	280	2.2967	0.5795	9.9413
2.4202	11.0	308	2.2407	0.5858	9.4003
2.3904	12.0	336	2.2537	0.5825	9.5226
2.3781	13.0	364	2.3002	0.5770	9.9764
2.3732	14.0	392	2.2511	0.5872	9.4982
2.3337	15.0	420	2.2408	0.5854	9.4009
2.3321	16.0	448	2.2484	0.5810	9.4722

Framework versions

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