Dani Bodor
Dani Bodor
As @cbaakman pointed out, the current feature module that calculates the nonbond energies in the "old" method is this: https://github.com/DeepRank/deeprank2/blob/main/deeprank2/features/contact.py
Note that test_trainer alsready has a `test_cuda` function. Not sure if this helps at all?
Maybe we can look for a way to check for NaN/missing values systematically across all features after generating the graph, and output an error with missing features and/or an option...
Good point about defaulting NaNs. I still think it would be a good idea to have a default check for NaNs during graph creation (maybe after each feature module is...
@gcroci2 , has this been addressed/solved yet?
Was tried in PR #454 , but didn't work. Might be something to do with apt (and conda?) installs not being cached properly. Feedback from Stef was: > ik denk...
@gcroci2, are we closing this issue?
I think it's not that straightforward. I think was working on this a while ago and got completely stuck and ended up having to ditch the entire branch, because it...
> > I think it's not that straightforward. I think was working on this a while ago and got completely stuck and ended up having to ditch the entire branch,...
also incorporate #608 in mkdocs?