Oh absolutely! The LLM equivalent of "death and taxes" is prefill and decode; and that holds true irrespective of proprietary inference optimizations. Less ver

Oh absolutely! The LLM equivalent of "death and taxes" is prefill and decode; and that holds true irrespective of proprietary inference optimizations. Less verbose output = less context & less token gen.