Skip to content

[MoE Refactor] Rename "naive" all2all backend - #36294

Merged
robertgshaw2-redhat merged 7 commits into
vllm-project:mainfrom
neuralmagic:remove-naive
Mar 19, 2026
Merged

[MoE Refactor] Rename "naive" all2all backend#36294
robertgshaw2-redhat merged 7 commits into
vllm-project:mainfrom
neuralmagic:remove-naive

Conversation

@bnellnm

@bnellnm bnellnm commented Mar 7, 2026 •
edited by github-actions Bot
Loading

Copy link
Copy Markdown
Collaborator

Purpose

Removenaive all2all backend option for the FusedMoE layer.

Test Plan

MoE refactoring tests.

Test Result

cc @robertgshaw2-redhat

Essential Elements of an Effective PR Description Checklist
  • The purpose of the PR, such as "Fix some issue (link existing issues this PR will resolve)".
  • The test plan, such as providing test command.
  • The test results, such as pasting the results comparison before and after, or e2e results
  • (Optional) The necessary documentation update, such as updating supported_models.md and examples for a new model.
  • (Optional) Release notes update. If your change is user facing, please update the release notes draft in the Google Doc.

bnellnm added 2 commits March 7, 2026 00:17
@bnellnm
remove naive all2all backend
74c5c7f 
Signed-off-by: Bill Nell <bnell@redhat.com>
@bnellnm
cleanup
c3062b8 
Signed-off-by: Bill Nell <bnell@redhat.com>
gemini-code-assist[bot]
gemini-code-assist Bot reviewed Mar 7, 2026
View reviewed changes

@gemini-code-assist gemini-code-assist Bot left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Code Review

This pull request correctly removes the 'naive' all2all backend option for MoE layers. The changes are consistent across the configuration and logic files, replacing the removed backend with a fallback and cleaning up its references. I have one suggestion to improve code clarity by renaming a property that has become a misnomer after this change.

Comment thread vllm/model_executor/layers/fused_moe/config.py Resolved
ProExpertProg
ProExpertProg approved these changes Mar 7, 2026
View reviewed changes

@ProExpertProg ProExpertProg left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Should we update the PR title to say "rename" instead of "remove"?

yewentao256
yewentao256 approved these changes Mar 7, 2026
View reviewed changes

@yewentao256 yewentao256 left a comment

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Should we explicitly tell users now allgather_reducescatter is used by default?

@yewentao256 yewentao256 added the ready ONLY add when PR is ready to merge/full CI is needed label Mar 9, 2026
hmellor
hmellor reviewed Mar 13, 2026
View reviewed changes
Comment thread vllm/config/parallel.py Resolved
@bnellnm bnellnm changed the title [MoE Refactor] Remove "naive" all2all backend [MoE Refactor] Rename "naive" all2all backend Mar 16, 2026
bnellnm added 2 commits March 16, 2026 19:00
@bnellnm
review comments
3448677 
Signed-off-by: Bill Nell <bnell@redhat.com>
@bnellnm
Merge remote-tracking branch 'nm-vllm/remove-naive' into remove-naive
0a39ad6 
Signed-off-by: Bill Nell <bnell@redhat.com>
@mergify mergify Bot added the nvidia label Mar 16, 2026
@github-project-automation github-project-automation Bot moved this to Ready in NVIDIA Mar 16, 2026
@bnellnm

bnellnm commented Mar 17, 2026

Copy link
Copy Markdown
Collaborator Author

Should we explicitly tell users now allgather_reducescatter is used by default?

Yeah, makes sense. Not really sure where to do that. Do you know if there are docs?

@bnellnm
remove references to naive from docs
128a45c 
Signed-off-by: Bill Nell <bnell@redhat.com>
@bnellnm

bnellnm commented Mar 17, 2026

Copy link
Copy Markdown
Collaborator Author

Should we explicitly tell users now allgather_reducescatter is used by default?

Yeah, makes sense. Not really sure where to do that. Do you know if there are docs?

I searched the docs and removed all references to naive. It looks like we already say that allgather_reducescatter is the default.

@mergify

mergify Bot commented Mar 17, 2026

Copy link
Copy Markdown
Contributor

Documentation preview: https://vllm--36294.org.readthedocs.build/en/36294/

@mergify mergify Bot added the documentation Improvements or additions to documentation label Mar 17, 2026
@robertgshaw2-redhat
robertgshaw2-redhat merged commit 91be5f9 into vllm-project:main Mar 19, 2026
60 of 61 checks passed
@github-project-automation github-project-automation Bot moved this from Ready to Done in NVIDIA Mar 19, 2026
chooper26 pushed a commit to intellistream/vllm-hust-legacy-20260831 that referenced this pull request Mar 21, 2026
@bnellnm @ShuhaoZhangTony
[MoE Refactor] Rename "naive" all2all backend (vllm-project#36294)
a5c0e75 
Signed-off-by: Bill Nell <bnell@redhat.com>
SouthWest7 pushed a commit to SouthWest7/vllm that referenced this pull request Mar 27, 2026
@bnellnm SouthWest7
[MoE Refactor] Rename "naive" all2all backend (vllm-project#36294)
63f1cf6 
Signed-off-by: Bill Nell <bnell@redhat.com>
khairulkabir1661 pushed a commit to khairulkabir1661/vllm that referenced this pull request Mar 27, 2026
@bnellnm @khairulkabir1661
[MoE Refactor] Rename "naive" all2all backend (vllm-project#36294)
2f237e9 
Signed-off-by: Bill Nell <bnell@redhat.com>
JiantaoXu pushed a commit to JiantaoXu/vllm that referenced this pull request Mar 28, 2026
@bnellnm @xujiantaoxjt-cyber
[MoE Refactor] Rename "naive" all2all backend (vllm-project#36294)
f0ce6f7 
Signed-off-by: Bill Nell <bnell@redhat.com>
mtparet pushed a commit to blackfuel-ai/vllm that referenced this pull request Apr 9, 2026
@bnellnm @mtparet
[MoE Refactor] Rename "naive" all2all backend (vllm-project#36294)
d55614c 
Signed-off-by: Bill Nell <bnell@redhat.com>
@bnellnm
bnellnm deleted the remove-naive branch April 15, 2026 20:49
mystous pushed a commit to mystous/vllm_hybrid that referenced this pull request May 10, 2026
@bnellnm
[MoE Refactor] Rename "naive" all2all backend (vllm-project#36294)
e5149e0 
Signed-off-by: Bill Nell <bnell@redhat.com>
my-other-github-account pushed a commit to my-other-github-account/vllm that referenced this pull request May 15, 2026
@bnellnm
[MoE Refactor] Rename "naive" all2all backend (vllm-project#36294)
f580e92 
Signed-off-by: Bill Nell <bnell@redhat.com>
@UgaTheDev UgaTheDev mentioned this pull request Jul 31, 2026
Open
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@hmellor hmellor hmellor left review comments
@yewentao256 yewentao256 yewentao256 approved these changes
@ProExpertProg ProExpertProg ProExpertProg approved these changes
@robertgshaw2-redhat robertgshaw2-redhat robertgshaw2-redhat approved these changes
@WoosukKwon WoosukKwon Awaiting requested review from WoosukKwon WoosukKwon is a code owner
@youkaichao youkaichao Awaiting requested review from youkaichao youkaichao is a code owner
@mgoin mgoin Awaiting requested review from mgoin mgoin is a code owner
@tlrmchlsmth tlrmchlsmth Awaiting requested review from tlrmchlsmth tlrmchlsmth is a code owner
@houseroad houseroad Awaiting requested review from houseroad houseroad is a code owner
@pavanimajety pavanimajety Awaiting requested review from pavanimajety pavanimajety is a code owner
+1 more reviewer
@gemini-code-assist gemini-code-assist[bot] gemini-code-assist[bot] left review comments
Reviewers whose approvals may not affect merge requirements

Assignees

No one assigned

Labels

documentation Improvements or additions to documentation nvidia ready ONLY add when PR is ready to merge/full CI is needed

Projects

NVIDIA
Status: Done

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

5 participants

@bnellnm @ProExpertProg @hmellor @yewentao256 @robertgshaw2-redhat