Comment by adeon

I think the LLaMA-7B in general might not just be very good. I've been playing around and run full non-quantized LLaMA-30B and LLaMA-7B in a bunch of experiments and I think the quality of output is much, much better in LLaMA-30B.