← Quit0 / 0 correct
LLMs & AIHard1 of 10

Prefill vs decode phases of LLM inference: which is accurate?