llama.cpp

4b51a5f2 - handle min/max token counts from CLI