llama.cpp

fa3582f5 - Tell users attmepting to run perplexity with too few tokens to use more (#2882)