llama.cpp

use fprintf for diagnostic output, keep printf only for printing model output

#48

Merged