Thread

started actually using my mac neural chip for a 70B parameter Llama model in building a private personal assistant and I'm at max RAM too quick