from Hacker News

Efficient Text-to-Image Training (16x cheaper than Stable Diffusion) [video]

by jhncls on 9/15/23, 3:52 PM with 1 comments

  • by billconan on 9/15/23, 4:53 PM

    I don't get it, if the efficient net can compress the image further, why not using the efficient net as the compressor? what's the point of using a vqgan?