llm.c is now down to 26.2ms/iteration, matching PyTorch by from Hacker News on 2024-04-14 08:58 (#6M2DA) Comments