정다혜

@twaka hi there, I've tried it myself and also observed the model weights reduced using Llama3-8B. But for me, although the memory consumption decreased, the latency seems to increase significantly...