llama.cpp

c0ea2b30 - rm get_prev_tokens, move it to the main pr