Learner Brightening/(Reducing Contrast) images while making prediction

I don’t understand how to explain this. But my model was running fine yesterday on Kaggle. When I re-run it today, every image looks like it’s approaching a brighter and brighter colour. I don’t understand why this is happening or what to do to stop it. I don’t think I’ve changed anything about it.

Here’s the Ouput data loader:

And here are today’s predictions:

This was Yesterday:

I’ve not changed anything between now and then. I just loaded it in the notebook and ran it again. This is what I get. Any clue as to why this is happening would be greatly helpful.

I am building a super-resolution model on Div2K using VGG-16 Perceptual Loss as a loss function.

I don’t know if this matters but this is how I am loading in the model:
It is a part of a data set I uploaded to Kaggle and I am taking it in from there.

Any help at all would be greatly appreciated.