llama.cpp

ggml : revert change to ggml_cpy, add ggml_cont_Nd instead

#3275

Merged