[Grimblewald]

as it stands Llama3 cant even produce sensible responses approaching its quoted max context, but if this increases the window we have before nonsense begins, then that is awesome!