CancerTextV2

This model is a fine-tuned version of bionlp/bluebert_pubmed_uncased_L-12_H-768_A-12 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.5913
• Accuracy: 0.8692
• Precision: 0.8666
• Recall: 0.8738
• F1: 0.8701

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: 1e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Precision	Recall	F1
0.4717	1.0	600	0.3318	0.8617	0.8562	0.8704	0.8633
0.3248	2.0	1200	0.3144	0.8658	0.8821	0.8455	0.8634
0.2653	3.0	1800	0.3519	0.8625	0.8507	0.8804	0.8653
0.2164	4.0	2400	0.4090	0.8658	0.9002	0.8239	0.8604
0.1884	5.0	3000	0.4413	0.8667	0.8850	0.8439	0.8639
0.1582	6.0	3600	0.4415	0.865	0.8971	0.8256	0.8599
0.1377	7.0	4200	0.5165	0.8708	0.8694	0.8738	0.8716
0.1192	8.0	4800	0.5699	0.8675	0.8826	0.8488	0.8654
0.1081	9.0	5400	0.5837	0.8692	0.8666	0.8738	0.8701
0.1018	10.0	6000	0.5913	0.8692	0.8666	0.8738	0.8701

Framework versions

• Transformers 4.21.2
• Pytorch 1.12.1+cu113
• Datasets 2.4.0
• Tokenizers 0.12.1