Skip to content

perf: avoid full-vocab all-gather for draft greedy sampling - #1534

Open
sufubao wants to merge 1 commit into
ModelTC:mainfrom
sufubao:exact-vocab-parallel-greedy
Open

perf: avoid full-vocab all-gather for draft greedy sampling#1534
sufubao wants to merge 1 commit into
ModelTC:mainfrom
sufubao:exact-vocab-parallel-greedy

Commits

Commits on Sep 1, 2026