add pytorch-geometric style batching for the online backend#84
Open
tbrugere wants to merge 1 commit into
Open
Conversation
This file contains hidden or 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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Hi!
This is pretty much WIP, but I added the option to use pytorch-geometric style batching, like mentioned in this comment: #35 (comment)
It is only added for the online backend (I think adding it to the multiscale backend would require a major rework)
Besides that, I am kind of worried that in tests, using this batching introduces a ~1% difference from computing the losses independently (see test_scrip.py)… I am not sure whether to attribute it to numerical error or if there is a subtle error I made in the implementation
Please have a look when you get the time!