New technique cuts AI model storage by 32x with minimal performance loss

Original: wait 32x storage reduction and almost no performance hit?! https://t.co/Fdi82KqgtP

Source: x.com ↗