CancerTextV3

This model is a fine-tuned version of dmis-lab/biobert-base-cased-v1.1 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.6660
• Accuracy: 0.8658
• Precision: 0.8448
• Recall: 0.8868
• F1: 0.8653

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
0.4603	1.0	600	0.3671	0.8575	0.8056	0.9314	0.8640
0.3208	2.0	1200	0.3296	0.8658	0.8613	0.8628	0.8620
0.2619	3.0	1800	0.4023	0.8667	0.8797	0.8405	0.8596
0.2125	4.0	2400	0.3948	0.8733	0.8459	0.9039	0.8740
0.1789	5.0	3000	0.4971	0.8617	0.8243	0.9091	0.8646
0.1568	6.0	3600	0.5372	0.8633	0.8384	0.8902	0.8636
0.1332	7.0	4200	0.6037	0.865	0.8502	0.8765	0.8632
0.1176	8.0	4800	0.6255	0.8675	0.8430	0.8937	0.8676
0.1043	9.0	5400	0.6479	0.865	0.8434	0.8868	0.8645
0.096	10.0	6000	0.6660	0.8658	0.8448	0.8868	0.8653

Framework versions

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