search

DeepGraphLearning / torchdrug

A powerful and flexible machine learning platform for drug discovery
https://torchdrug.ai/
Apache License 2.0
1.42k stars 200 forks source link

A problem with Retrosynthesis #102

Closed lemon-prog123 closed 2 years ago

lemon-prog123 commented 2 years ago

Hi, I have build a pipline with your guideline in Retrosynthesis Moudle. It works well on default dataset. But when I tried to test it on my own smiles, there was something wrong with it. To be specific, I convert my smiles with data.Molecule.from_smiles(smile,node_feature="synthon_completion", kekulize=True) just as test_set do. But it seems that the molecule in test set has 43 types of nodes. But my converted molecule from smiles only has 39 types of nodes. So it can't be sent to the model as the input-dims is 43.

How can I fix it?

For example,my smile "C=C(CC)C(C)(C)C"

zqsaudience commented 11 months ago

Hi，Can I find out how you solved this problem?