Someone needs to scale up Llama 2 to GPT 3 size. Would be curious how that compares then.

→ View original post on X — @rasbt