Z zhihaojia.medium.com Compiling LLMs into a MegaKernel: A path to low-latency inference 314 points matt_d 1 year ago 76 comments
