[vLLM][moe] Route blockwise FP8 fused MoE to Triton#7569
Draft
sarah12121212 wants to merge 2 commits into
Draft
Conversation
sarah12121212
changed the base branch from
main
to
moe2-workspacemanager-xpu-init
July 24, 2026 14:20
sarah12121212
force-pushed
the
moe2-fp8-moe
branch
from
July 24, 2026 14:21
57707f8 to
e5b996c
Compare
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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
For blockwise quantization key pair
activation_key == kFp8Dynamic128Sym and weight_key == kFp8Static128BlockSymroute fused MoE through Triton instead of the native Xe2 grouped-GEMM path, which accepts only 1D per-expert scales and rejects multidimensional scale grids.Unskip the 44 affected Xe2 tests.
Blocked by #7462 (it needs to be merged first)
Relates to issue #7574
Testing
vLLM MoE suite on B580 (tested locally):
vLLM MoE suite on PVC max 1100: