[DNM] add experiment of training torsion energies #163
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
This is amazing!! cheers. I'll dive into it now. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Do not merge: this is only a very partial, bare-bones experiment to see how much work it would be to use NAGL to a) train to non-atom properties (not so bad) and b) incorporate geometric information (more work...) and c) incorporate another trained parameter (quite a lot). My general conclusion is that b) and c) is likely not worth porting to the overall library, although open to anything as usual.
All the maths is very ad-hoc and absolutely needs double-checking and testing, especially the geometry functions. Everything currently is only implemented with the DGL backend.
cc @BenCree -- there's an example notebook here that runs through a very short example of what I think you and Danny were describing to me in our call, if you're interested. Happy to chat more if you want to discuss!
P.S. the Dataset code was a huge pain and very repetitive -- it probably needs refactoring.
PR Checklist