Comment by simonw

Does the chinchilla recipe still hold today? I got the impression that the LLaMA paper proposed a different result where throwing far more tokens at the problem had a very meaningful impact, or did I misunderstand that?