Randolph-zeng

> @minsis are you using on openshift by any chance :) lol, my openshift pod is failing to start while if I use `oc debug` into the pod it runs...

Just curious, am I the only one that experienced the DDIM degradation here ? When I use the code in this PR I noticed that the DDIM almost failed completely...

There is a ICLR 2022 paper "PROMPT-TO-PROMPT IMAGE EDITING WITH CROSS-ATTENTION CONTROL" that does similar but more things into the pipeline. There are also open sourced code in the web,...

> > "the coefficient of epsilon is not sqrt-ed": > > It is in the original repo: [this line](https://github.com/andreas128/RePaint/blob/8e2a0552a0cc306e623da807d353915f1af17726/guided_diffusion/gaussian_diffusion.py#L368) But you are right the coefficient square root is missing from...

btw, It seems tha there are some errors regarding compatability and formatting, if someone can confirm that this is indeed an undesired off by one error I am happy to...

oh yeah, thanks for the catch! I will incorporate that and fix the errors by doing some local test

Hi @patrickvonplaten , I totally agree with you that SD-Legacy is not a standard pipeline following any paper and is different from RePaint. It is just that every diffusion models...

For the most libs I believe they have some extent of similarity(e.g. TRL, TRLX, ClossalAI, RL4LM etc), so I am not sure if they are referencing your code or more...