llama : overlap the AllReduce of split mode tensor with the compute of the next ubatch - #17

llama : overlap the AllReduce of split mode tensor with the compute of the next ubatch#17

Piggidragon merged 3 commits into

mainPiggidragon/llama.cpp:mainfrom

llama/split-overlapPiggidragon/llama.cpp:llama/split-overlapCopy head branch name to clipboard