Repository navigation

Speed up M=4 packed GEMV and multi-row lm_head for NEXTN - #1

Merged

krisclarkdev merged 1 commit into

gptq-page-sizefrom

gptq-decode-speed

Oct 29, 2026

Merged

Speed up M=4 packed GEMV and multi-row lm_head for NEXTN#1
krisclarkdev merged 1 commit into
gptq-page-sizefrom
gptq-decode-speed

Commits on Oct 18, 2026
- Speed up M=4 packed GEMV and multi-row lm_head for NEXTN decode.
  
  krisclarkdev
  and
  cursoragent
  committed
  17d0385