nopainkiller/Qwen1.5-4x0.5B-MoE

Of course this is a unworkable model, I only use this to test mergekit and llama.cpp on my Apple M2 Macbook, do not use unless I somehow finetune it with the MoD dataset. My minor mergekit fix for anyone using on M2 with Rossetta.