llama.cpp

llama : fix command-r inference when omitting outputs

#6367

Merged