llama.cpp

Tell users attmepting to run perplexity with too few tokens to use more

#2882

Merged