papersnake·4 वर्ष पहले·discussHave you tested this on big models involving multi-gpu communication, or any plans?