qwen3_5 cuda: map an EXL3 embedding from host memory only when the card has no room - #1678

qwen3_5 cuda: map an EXL3 embedding from host memory only when the card has no room#1678

ashhart merged 2 commits into

ashhart:mainashhart/TensorFold:mainfrom

diwakergupta:qwen35-exl3-host-embeddiwakergupta/TensorFold:qwen35-exl3-host-embedCopy head branch name to clipboard