vllm
[Quantization] Bound peak memory when repacking FP4 MoE weights for Marlin
#47851

Merged

[Quantization] Bound peak memory when repacking FP4 MoE weights for Marlin #47851

mgoin merged 3 commits into vllm-project:main from joerowell:joerowell/fix_marlin_overuse

[Quantization] Bound peak memory when repacking FP4 MoE weights for M…

6924642d

joerowell requested a review from

mgoin 115 days ago

joerowell requested a review from

robertgshaw2-redhat 115 days ago

joerowell requested a review from

tlrmchlsmth 115 days ago

joerowell requested a review from

yewentao256 115 days ago

joerowell requested a review from

pavanimajety 115 days ago

joerowell requested a review from

zyongye 115 days ago

claude commented on 2026-07-07

Simplify

75604c44

mgoin added bug

mgoin added ready

mgoin added nvidia

Merge branch 'main' into joerowell/fix_marlin_overuse

e858f063

mgoin enabled auto-merge (squash) 112 days ago

mgoin approved these changes on 2026-07-11

mgoin merged 1bf3997e into main 112 days ago

Reviewers

mgoin

claude

robertgshaw2-redhat

tlrmchlsmth

yewentao256

pavanimajety

zyongye

Assignees

No one assigned

Labels

bug ready nvidia

Milestone

No milestone