3 ms·One Layer Is Enough: Adapting Pretrained Visual Encoders for Image Generation2 points by badmonster 10mo ago