CeLLaTe-ner-2class-tapt-pubmedbert-tokenizer-original-spanmask

This model is a fine-tuned version of Mardiyyah/CeLLaTe-tapt-pubmedbert-tokenizer-original-spanmask on the OTAR3088/CeLLaTe-ner-2class-iob_final dataset. It achieves the following results on the evaluation set:

• Loss: 0.0797
• Precision: 0.7923
• Recall: 0.7285
• Micro F1: 0.7591
• Weighted F1: 0.7590
• Macro F1: 0.7671
• Accuracy: 0.9838

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: 2e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 3407
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.01
• num_epochs: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	Micro F1	Weighted F1	Macro F1	Accuracy
0.2444	1.0	263	0.0630	0.5922	0.7267	0.6526	0.6546	0.6663	0.9787
0.0414	2.0	526	0.0545	0.7804	0.7157	0.7467	0.7467	0.7529	0.9834
0.0262	3.0	789	0.0608	0.7595	0.7285	0.7437	0.7439	0.7511	0.9832
0.0198	4.0	1052	0.0625	0.7483	0.7383	0.7432	0.7436	0.7507	0.9836
0.014	5.0	1315	0.0738	0.7883	0.7157	0.7502	0.7503	0.7588	0.9834
0.0114	6.0	1578	0.0727	0.7590	0.7187	0.7383	0.7386	0.7465	0.9831
0.0088	7.0	1841	0.0766	0.7602	0.7407	0.7503	0.7505	0.7602	0.9836
0.0066	8.0	2104	0.0795	0.7916	0.7279	0.7584	0.7584	0.7667	0.9838
0.0055	9.0	2367	0.0845	0.7180	0.7736	0.7448	0.7459	0.7581	0.9829
0.0047	10.0	2630	0.0895	0.7877	0.7175	0.7510	0.7510	0.7554	0.9839
0.0038	11.0	2893	0.0861	0.7614	0.7261	0.7433	0.7436	0.7509	0.9842
0.0033	12.0	3156	0.1008	0.7880	0.7053	0.7444	0.7444	0.7563	0.9836
0.0027	13.0	3419	0.0966	0.7549	0.7309	0.7427	0.7429	0.7474	0.9834

Framework versions

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