Run submodel inference in parallel using CUDA streams#18
Merged
Conversation
mrwnmsr
approved these changes
Jun 10, 2026
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
3 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
So far, we've been running RetroChimera's submodels sequentially. Since they are independent, we can run them in parallel; this has potential to speed things up if the submodels are not fully utilizing the GPU, which is the case for the ones currently used in RetroChimera. This PR adds an option to run the submodels in parallel using CUDA streams. This is hidden behind a flag, which is
Trueby default (the main reason to turn it off would be if operating under very stringent memory requirements, as parallelization does slightly increase memory needs).The speedup is small but detectable, and model output is unchanged. For context, I also include a comparison with the state of
mainbefore #15 (back from before we started looking at model speed altogether).