Is noise optimized in the first step?

[caopulan]

PTI/training/projectors/w_projector.py

Line 47 in fb1c485

G = copy.deepcopy(G).eval().requires_grad_(False).to(device).float() # type: ignore

Although 'noise constant' is updated at the end of optizimation, G is a deepcopy version and deleted. So does optimized noise work in inference?

[rardz]

From the code, it seems the optimized 'noise constant' in the first step will be abandoned, and in the next pti step, the model just use another irrelevant constant noise.
However it was a little strange why the authors choose to keep the pti step noise constant. Especially in the multi image pti case, how will different images should share a same set of constant noise? I also think there must be something wrong here.