1.58-Bit Fluxarxiv.org 2 pointsreynaldi1 year ago1 commentSaveHideCopy link On HNComments−reynaldiOP1ySome impressive results1.58-bit FLUX achieves a 7.7× reduction in model storage and more than a 5.1× reduction in inference memory usage
Comments
Some impressive results