Correctly calculate distributed loss average #269
Merged
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
abhshkdz
approved these changes
Sep 18, 2021
sparticlesteve
pushed a commit
to sparticlesteve/ocp
that referenced
this pull request
May 23, 2022
Co-authored-by: Abhishek Das <das.abhshk@gmail.com>
This was referenced May 23, 2022
levineds
pushed a commit
that referenced
this pull request
Jul 11, 2024
Co-authored-by: Abhishek Das <das.abhshk@gmail.com>
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Outdated suggestions cannot be applied.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.
We previously first calculated the loss average per DDP replica and then averaged across replicas. This leads to wrong results since not every replica has the same number of atoms or even systems. This PR changes the computation to instead calculate an average over all replicas by aggregating the number of samples.
This seems to give a tiny improvement in MAE (0.5% or so). But training looks very similar to previously.