▲ 3 pointsAutoMegaKernel: Compiling a LLM into a single CUDA kernelarxiv.orgby OsamaJaber·2mo ago·0 comments·view on hn ↗