Image Oversaturation/Image Corruption by using denoising process

I know the authors aren't responding to issues, so maybe some of the users might have some insight...

Does the original code written in tensorflow result in corrupt images? Mainly images with high contrast, such as half white/half black image...

I have executed converted pytorch code and images suffer from corruption and artifacts especially at regions that has high contrast...

Thanks