News for hackers. Stuff that matters.

papersnake

Have you tested this on big models involving multi-gpu communication, or any plans?