dflash: zero-fill draft-cache holes left by mtmd chunks and reused prefixes - #1

dflash: zero-fill draft-cache holes left by mtmd chunks and reused prefixes#1

dagnarf wants to merge 2 commits into

z-lab:dflash2z-lab/llama.cpp-fork:dflash2from

dagnarf:dflash2-mtmd-zero-filldagnarf/llama.cpp-fork:dflash2-mtmd-zero-fillCopy head branch name to clipboard