vllm
[Perf][MXFP4] Optimize activation quant for mxfp4. 3x kernel speedup, 5% throughput improvement
#63101
Merged
Go
Login via GitHub
Home
Pricing
FAQ
Install
Login
via GitHub
Overview
Commits
9
Changes
View On
GitHub
[Perf][MXFP4] Optimize activation quant for mxfp4. 3x kernel speedup, 5% throughput improvement
#63101
mgoin
merged 9 commits into
vllm-project:main
from
neuralmagic:imarkov/mxfp4-optimize-activation-conversion
mergify
added
ci/build
mergify
added
performance
mergify
added
quantization
mergify
added
nvidia
Use sorted_experts_ids to avoid the search
2b96154f
ilmarkov
force pushed
from
9ef3ff92
to
2b96154f
30 days ago
Address comments
798e259a
cleanup
00b313c9
ilmarkov
marked this pull request as ready for review
30 days ago
ilmarkov
requested a review
from
bnellnm
30 days ago
ilmarkov
requested a review
from
mgoin
30 days ago
ilmarkov
requested a review
from
pavanimajety
30 days ago
ilmarkov
requested a review
from
zyongye
30 days ago
ilmarkov
requested a review
from
wzhao18
30 days ago
claude
commented on 2026-10-22
Update tests and benchmarks
519f1857
LopezCastroRoberto
commented on 2026-10-22
LopezCastroRoberto
commented on 2026-10-22
yewentao256
commented on 2026-10-22
Address comments. Cleanup
6c8a3892
Upd benchmark
c2c88636
LopezCastroRoberto
approved these changes on 2026-10-23
LopezCastroRoberto
added
ready
Merge branch 'main' into imarkov/mxfp4-optimize-activation-conversion
e7ea1a93
Merge branch 'main' into imarkov/mxfp4-optimize-activation-conversion
992108ff
Merge branch 'main' into imarkov/mxfp4-optimize-activation-conversion
21ae9b68
mgoin
approved these changes on 2026-10-26
mgoin
merged
f3bdb0e0
into main
25 days ago
Login to write a write a comment.
Login via GitHub
Reviewers
mgoin
LopezCastroRoberto
yewentao256
claude
bnellnm
pavanimajety
zyongye
wzhao18
Assignees
No one assigned
Labels
performance
ready
ci/build
nvidia
quantization
Milestone
No milestone
Login to write a write a comment.
Login via GitHub