DiffSBDD icon indicating copy to clipboard operation
DiffSBDD copied to clipboard

Published 20 hours ago • verified publisher icon arneschneuing

High CUDA memory usage while generating ligands

Open CADDfarmer opened this issue 1 year ago • 0 comments

Hi all,

When I ran generate_ligands.py with parameters "--sanitize --relax --num_nodes_lig 50 --n_samples 20" in moad_fullatom_cond model. It took 5368MB CUDA memory!!!

Is it OK for only sampling 20 molecules with high CUDA memory usage? And, how can I generate 10 thousand or more molecules?

CADDfarmer avatar Oct 20 '23 01:10 CADDfarmer