Training Text-to-Image Models Without a VAE

by schopra909 | View on Hacker News