Comment by mitxela
3 days ago
Well, you don't. That's why 1.58-bit (ternary) quants are often used. But if that's what they're going for, no need to dress it up in a floating-point facade.
3 days ago
Well, you don't. That's why 1.58-bit (ternary) quants are often used. But if that's what they're going for, no need to dress it up in a floating-point facade.
No comments yet
Contribute on Hacker News ↗