vllm
[Perf][MXFP4] Optimize activation quant for mxfp4. 3x kernel speedup, 5% throughput improvement
#63101
Merged

[Perf][MXFP4] Optimize activation quant for mxfp4. 3x kernel speedup, 5% throughput improvement #63101

ilmarkov
mergify mergify added ci/build
mergify mergify added performance
mergify mergify added quantization
mergify mergify added nvidia
ilmarkov Use sorted_experts_ids to avoid the search
2b96154f
ilmarkov ilmarkov force pushed from 9ef3ff92 to 2b96154f 30 days ago
ilmarkov Address comments
798e259a
ilmarkov cleanup
00b313c9
ilmarkov ilmarkov marked this pull request as ready for review 30 days ago
ilmarkov ilmarkov requested a review from bnellnm bnellnm 30 days ago
ilmarkov ilmarkov requested a review from mgoin mgoin 30 days ago
ilmarkov ilmarkov requested a review from pavanimajety pavanimajety 30 days ago
ilmarkov ilmarkov requested a review from zyongye zyongye 30 days ago
ilmarkov ilmarkov requested a review from wzhao18 wzhao18 30 days ago
claude
claude commented on 2026-10-22
ilmarkov Update tests and benchmarks
519f1857
LopezCastroRoberto
LopezCastroRoberto commented on 2026-10-22
LopezCastroRoberto
LopezCastroRoberto commented on 2026-10-22
yewentao256
yewentao256 commented on 2026-10-22
ilmarkov Address comments. Cleanup
6c8a3892
ilmarkov Upd benchmark
c2c88636
LopezCastroRoberto
LopezCastroRoberto approved these changes on 2026-10-23
LopezCastroRoberto LopezCastroRoberto added ready
LopezCastroRoberto Merge branch 'main' into imarkov/mxfp4-optimize-activation-conversion
e7ea1a93
github-actions
LopezCastroRoberto
github-actions
LopezCastroRoberto
ilmarkov Merge branch 'main' into imarkov/mxfp4-optimize-activation-conversion
992108ff
LopezCastroRoberto
github-actions
LopezCastroRoberto Merge branch 'main' into imarkov/mxfp4-optimize-activation-conversion
21ae9b68
LopezCastroRoberto
github-actions
mgoin
mgoin approved these changes on 2026-10-26
mgoin mgoin merged f3bdb0e0 into main 25 days ago

Login to write a write a comment.

Login via GitHub

Assignees
No one assigned
Labels
Milestone