Why does the pt file have inconsistent parameters?

[AOSAI]

I used guided-diffusion's image_train.py to train my own dataset, although the training also encountered a lot of bugs, but finally started to do the training successfully.

What puzzles me is why there is such a thing as inconsistent parameters in the trained pt file? It makes me feel like there's nothing I can do about it at all.