[P] Using LoRA to efficiently fine-tune diffusion models. Output model less than 4MB, two times faster to train, with better performance. (Again, with Stable Diffusion) Submitted by cloneofsimo t3_zfkqjh on December 8, 2022 at 1:27 AM in MachineLearning 26 comments 114
[P] Implementation of MagicMix from ByteDance researchers, - New way to interpolate concepts with much more natural, geometric coherency (implemented with Stable Diffusion!) Submitted by cloneofsimo t3_ykiuq0 on November 2, 2022 at 9:54 PM in MachineLearning 15 comments 88