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
473 stars 109 forks source link

Potential improvement through avoiding DGL TensorDispatcher error #12

Closed PatWalters closed 2 years ago

PatWalters commented 2 years ago

I encountered this error when running Equibind.

/opt/dgl/src/runtime/tensordispatch.cc:43: TensorDispatcher: dlopen failed: libtorch_cuda.so: cannot open shared object file: No such file or directory

According to this https://discuss.dgl.ai/t/error-tensordispatcher/2468 the error impacts performance but does not impact usage. Equibind appears to run in spite of this error. I thought you might want to take a look.

Ubuntu 18.04.6 LTS Installed according to the readme with conda env create -f environment.yml

HannesStark commented 2 years ago

Thanks! We are investigating this and will keep you updated in this issue.

HannesStark commented 2 years ago

I was able to reproduce it on one machine. With the error being thrown the training time was the same as without. I am pretty sure that this does not impact runtime here. I am closing this issue but am happy about additional insights from anyone else!