fix: disable rope_fusion when context parallelism (cp > 1) is enabled#1440
Merged
fix: disable rope_fusion when context parallelism (cp > 1) is enabled#1440
Conversation
hemildesai
requested review from
HuiyingLi,
ZhiyuLi-Nvidia,
adil-a and
akoumpa
as code owners
hemildesai
force-pushed
the
worktree-disable-rope-fusion-cp
branch
from
ba3e60b to
f1edde7
Compare
Contributor
Author
|
/ok to test f1edde7 |
Fused RoPE is incompatible with context parallelism (cp > 1). This adds a guard in both the LLM and VLM recipe setup() methods to automatically set rope_fusion=False before model construction when cp_size > 1. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com> Signed-off-by: Hemil Desai <hemild@nvidia.com>
hemildesai
force-pushed
the
worktree-disable-rope-fusion-cp
branch
from
f1edde7 to
24dc2ad
Compare
Contributor
Author
|
/ok to test 24dc2ad |
SwekeR-463
pushed a commit
to SwekeR-463/Automodel
that referenced
this pull request
Mar 11, 2026
…NVIDIA-NeMo#1440) Fused RoPE is incompatible with context parallelism (cp > 1). This adds a guard in both the LLM and VLM recipe setup() methods to automatically set rope_fusion=False before model construction when cp_size > 1. Signed-off-by: Hemil Desai <hemild@nvidia.com> Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com> Signed-off-by: SwekeR-463 <swekerswasti@gmail.com>
linnanwang
pushed a commit
that referenced
this pull request
Apr 24, 2026
…#1440) Fused RoPE is incompatible with context parallelism (cp > 1). This adds a guard in both the LLM and VLM recipe setup() methods to automatically set rope_fusion=False before model construction when cp_size > 1. Signed-off-by: Hemil Desai <hemild@nvidia.com> Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
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.
Summary
Related to #1439
rope_fusion) is broken when context parallelism (cp_size > 1) is enabled. This PR adds a guard in both the LLM (train_ft.py) and VLM (finetune.py) recipesetup()methods to automatically disablerope_fusionbefore model construction whencp_size > 1.🤖 Generated with Claude Code