llama.cpp

125835b2 - server : refactor slot input data, move tokenizer to HTTP thread