Open SejeongPark8354 opened 1 year ago
Thanks for your interest in the work. Some parts of the algorithm use multiple CPUs , e.g. calculation of the KNN graph edges. However, actually we found that the sampling also worked well if we restricted it only on one CPU using the command taskset -c 0 python samplexxx.py
. It seemed that the number of CPUs did not influence sampling too much. So we actually recommend using one CPU for sampling.
First of all, thank you for sharing your wonderful research. After I run the Pocket2Mol, I have a question about CPU usage. I think that Pocket2Mol doesn't require large memory in GPU but uses CPU a lot. Is there any reason for specifying Pocket2Mol use CPU a lot rather than GPU? Wouldn't it cause a problem if I change the code to run process mainly using GPU?