vllm
[Model][Perf] Fuse Beam's TP all-reduce with its sandwich norms and MoE finalize
#62366
Merged

[Model][Perf] Fuse Beam's TP all-reduce with its sandwich norms and MoE finalize #62366

WoosukKwon merged 6 commits into main from woosuk/beam-fused-allreduce
WoosukKwon
mergify mergify added ci/build
mergify mergify added nvidia
WoosukKwon WoosukKwon force pushed from f1d7247e to 32fc9c80 33 days ago
Base automatically changed from woosuk/beam-model to main 33 days ago
mergify
WoosukKwon [Model][Perf] Fuse Beam's TP all-reduce with its sandwich norms and M…
4fb4e753
mergify mergify added documentation
mergify mergify added new-model
mergify mergify added speculative-decoding
mergify mergify added dflash
mergify mergify added mrv2
mergify
mergify mergify added needs-rebase
WoosukKwon WoosukKwon force pushed from 32fc9c80 to 4fb4e753 33 days ago
mergify mergify removed needs-rebase
WoosukKwon [Model][Perf] Build Beam's fused all-reduce ahead of time under csrc/…
5d683553
WoosukKwon [Model][Perf] Note FlashInfer v0.7.2 as the vendored MNNVL source
3ec2cf25
WoosukKwon [Model][Perf] Keep Beam's deferred MoE finalize TP-only
9fdc9ed5
WoosukKwon [Build] Gate Beam's csrc kernels under one CMake switch
df583a4b
WoosukKwon WoosukKwon marked this pull request as ready for review 30 days ago
WoosukKwon WoosukKwon requested a review from DarkLight1337 DarkLight1337 30 days ago
WoosukKwon WoosukKwon requested a review from ywang96 ywang96 30 days ago
WoosukKwon WoosukKwon requested a review from AndreasKaratzas AndreasKaratzas 30 days ago
WoosukKwon WoosukKwon requested a review from tlrmchlsmth tlrmchlsmth 30 days ago
WoosukKwon WoosukKwon requested a review from LucasWilkinson LucasWilkinson 30 days ago
WoosukKwon WoosukKwon requested a review from Harry-Chen Harry-Chen 30 days ago
WoosukKwon Merge branch 'main' into woosuk/beam-fused-allreduce
73195883
claude
claude commented on 2026-10-21
WoosukKwon WoosukKwon added ready
WoosukKwon
github-actions
WoosukKwon WoosukKwon merged e4e6bcaa into main 30 days ago
WoosukKwon WoosukKwon deleted the woosuk/beam-fused-allreduce branch 30 days ago

Login to write a write a comment.

Login via GitHub

Assignees
No one assigned
Labels
Milestone