llama.cpp

sampling : support multiple outputs per sequence

#19833

Open