[Refactor] Split prepare_attn into a reduction step and per-quant-format residual steps - #41084

[Refactor] Split prepare_attn into a reduction step and per-quant-format residual steps#41084

ch-wan merged 1 commit into

mainsgl-project/sglang:mainfrom

cheng/refactor/attention-input-boundarysgl-project/sglang:cheng/refactor/attention-input-boundaryCopy head branch name to clipboard