feat: Flash Attention and modernized KV Cache layout for Slow-AR - #40

feat: Flash Attention and modernized KV Cache layout for Slow-AR#40

Skyrion9 wants to merge 3 commits into

rodrigomatta:mainrodrigomatta/s2.cpp:mainfrom

Skyrion9:FASkyrion9/s2.cpp:FACopy head branch name to clipboard