ClinicalTextV4

This model is a fine-tuned version of emilyalsentzer/Bio_ClinicalBERT on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.5609
• Accuracy: 0.8658
• Precision: 0.8371
• Recall: 0.8939
• F1: 0.8646

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
0.4824	1.0	600	0.3630	0.8458	0.825	0.8609	0.8426
0.3314	2.0	1200	0.3583	0.8558	0.8252	0.8870	0.8550
0.2673	3.0	1800	0.3437	0.8583	0.8189	0.9043	0.8595
0.2255	4.0	2400	0.3678	0.8675	0.8302	0.9096	0.8680
0.1883	5.0	3000	0.4002	0.8642	0.8259	0.9078	0.8650
0.1562	6.0	3600	0.4695	0.8633	0.8352	0.8904	0.8620
0.1372	7.0	4200	0.4940	0.8658	0.8371	0.8939	0.8646
0.1269	8.0	4800	0.5376	0.865	0.8402	0.8870	0.8629
0.1097	9.0	5400	0.5539	0.8633	0.8397	0.8835	0.8610
0.0997	10.0	6000	0.5609	0.8658	0.8371	0.8939	0.8646

Framework versions

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