search

HannesStark / EquiBind

EquiBind: geometric deep learning for fast predictions of the 3D structure in which a small molecule binds to a protein
MIT License
477 stars 109 forks source link

Memory requirements for training? #29

Closed neild0 closed 2 years ago

neild0 commented 2 years ago

I've been trying to train the model using the RDKitCoords_flexible_self_docking.yml configuration, and after modifying the code based on #6, I am having a RAM overflow problem.

It seems like the code loads receptors into memory — just wondering if this is supposed it happen and take so much memory, and, if so, how much is recommended to run this project. Thanks!

HannesStark commented 2 years ago

I am not sure how much RAM exactly is needed to perform the preprocessing of the data. If you do not have sufficient RAM you can modify the code to preprocess the data in portions instead of doing everything at once. In pdbbind.py you find the preprocess function which currently loads all complexes into memory at the same time. You could instead do this for only e.g. 500 complexes at a time.