llama.cpp

llama : make tensor_split ptr instead of array

#2272

Merged