llama.cpp

llama : remove ggml_cont where possible

#14568

Merged