AutoMegaKernel: Compiling a LLM into a single CUDA kernelarxiv.org3 points·OsamaJaber··0 commentsOpen articleSaveView on HN