BERT-SMILES

This is graduation project, molecular property prediction(QSAR).

SMILES means that change molecules to strings using ASCII codes.
But, this methods have difficulty in identifying molecular similarity.
So, we adds molecular similarity information using tanimoto coefficient(SSP) and applys to BERT Model because SMILES is strings.

Setting

we access Yonsei server using mobaxterm and works on local vscode using ssh connect.

Pre-traing

fine-tunning