Comment on Ternary Bonsai: Top Intelligence at 1.58 BitsComments−swiftcoder4moDoes this sort of thing scale? Would a 30B or higher model see similar performance/memory gains under this scheme?
Comments
Does this sort of thing scale? Would a 30B or higher model see similar performance/memory gains under this scheme?