Modelpublic

moro01525/ICU_Returns_COReClinicalBioBERT

sourceHugging Faceupdated 3y agoView on Hugging Face

0likes5downloads

Model Card

ICUReturnsCOReClinicalBioBERT

This model is a fine-tuned version of bvanaken/CORe-clinical-outcome-biobert-v1 on an unknown dataset. It achieves the following results on the evaluation set:

—Loss: 1.8391
—F1:: 0.7210
—Roc Auc: 0.7335
—Precision with 0:: 0.9048
—Precision with 1:: 0.6641
—Recall with 0:: 0.5220
—Recal with 1:: 0.9451
—Accuracy:: 0.7335

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: 0.0001
—trainbatchsize: 32
—evalbatchsize: 16
—seed: 42
—optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
—lrschedulertype: linear
—num_epochs: 13

Training results

Training Loss	Epoch	Step	Validation Loss	F1:	Roc Auc	Precision with 0:	Precision with 1:	Recall with 0:	Recal with 1:	Accuracy:
No log	1.0	46	0.6908	0.4103	0.5330	0.875	0.5172	0.0769	0.9890	0.5330
No log	2.0	92	0.6839	0.4361	0.5357	0.7241	0.5194	0.1154	0.9560	0.5357
No log	3.0	138	0.7113	0.4827	0.5549	0.7174	0.5314	0.1813	0.9286	0.5549
No log	4.0	184	0.6089	0.6674	0.6703	0.7095	0.6435	0.5769	0.7637	0.6703
No log	5.0	230	0.6138	0.6533	0.6731	0.8316	0.6171	0.4341	0.9121	0.6731
No log	6.0	276	0.6892	0.7153	0.7253	0.8596	0.664	0.5385	0.9121	0.7253
No log	7.0	322	1.0376	0.6385	0.6703	0.9189	0.6069	0.3736	0.9670	0.6703
No log	8.0	368	1.1796	0.7088	0.7225	0.8932	0.6552	0.5055	0.9396	0.7225
No log	9.0	414	1.0800	0.7749	0.7802	0.9048	0.7143	0.6264	0.9341	0.7802
No log	10.0	460	2.0318	0.6717	0.6951	0.9176	0.6272	0.4286	0.9615	0.6951
0.3613	11.0	506	1.9762	0.6796	0.7005	0.9101	0.6327	0.4451	0.9560	0.7005
0.3613	12.0	552	1.7367	0.7469	0.7555	0.9043	0.6867	0.5714	0.9396	0.7555
0.3613	13.0	598	1.8391	0.7210	0.7335	0.9048	0.6641	0.5220	0.9451	0.7335

Framework versions

—Transformers 4.34.0
—Pytorch 2.1.0+cu121
—Datasets 2.14.5
—Tokenizers 0.14.1