HealthNLP/pubmedbert_conmod

test

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

• Loss: 0.6886
• Accuracy: 0.8143
• F1: [0.92816572 0.56028369 0.1 0.2633452 ]

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: 32
• eval_batch_size: 64
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• num_epochs: 10.0

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1
No log	1.0	37	0.4891	0.8235	[0.91702786 0.33333333 0. 0.10837438]
No log	2.0	74	0.4762	0.8321	[0.93139159 0.48466258 0. 0.22857143]
No log	3.0	111	0.5084	0.8208	[0.92995725 0.44887781 0. 0.19266055]
No log	4.0	148	0.5519	0.8105	[0.92421691 0.44444444 0.06557377 0.30769231]
No log	5.0	185	0.5805	0.8294	[0.93531353 0.52336449 0.09345794 0.27131783]
No log	6.0	222	0.6778	0.7955	[0.91344509 0.55305466 0.15463918 0.29166667]
No log	7.0	259	0.6407	0.8213	[0.93298292 0.51383399 0.10191083 0.2519084 ]
No log	8.0	296	0.6639	0.8272	[0.9326288 0.55052265 0.18181818 0.26271186]
No log	9.0	333	0.6863	0.8192	[0.93071286 0.55830389 0.11042945 0.2761194 ]
No log	10.0	370	0.6886	0.8143	[0.92816572 0.56028369 0.1 0.2633452 ]

Framework versions

• Transformers 4.37.2
• Pytorch 2.2.0+cu121
• Datasets 2.17.0
• Tokenizers 0.15.2