Allocate tensors directly on target device#654
Open
GMNGeoffrey wants to merge 1 commit intojwohlwend:mainfrom
Open
Allocate tensors directly on target device#654GMNGeoffrey wants to merge 1 commit intojwohlwend:mainfrom
GMNGeoffrey wants to merge 1 commit intojwohlwend:mainfrom
Conversation
I was hitting a weird crash when running Boltz1 with DDP on ROCm that I narrowed down to one of these. Allocating directly on device seems to fix it, and is also cleaner and a minor performance optimization, so I went through and updated all the code to do this (obviously may have missed something).
volgin
added a commit
to Novel-Therapeutics/boltz-community
that referenced
this pull request
Mar 8, 2026
Avoids unnecessary CPU allocation + device transfer for tensors that are immediately moved to GPU/MPS/ROCm. Fixes a crash on ROCm with DDP. Inspired by upstream PR jwohlwend#654. Co-Authored-By: Claude Opus 4.6 <[email protected]>
|
We applied the fixes from this PR in boltz-community, a community-maintained fork with 20+ upstream bug fixes, broader hardware compatibility (MPS, ROCm), expanded test coverage, and CI. One of the changes ( |
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
2 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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
I was hitting a weird crash when running Boltz1 with DDP on ROCm that I narrowed down to one of these. Allocating directly on device seems to fix it, and is also cleaner and a minor performance optimization, so I went through and updated several other places as well (obviously may have missed some).