Improve predictor module

[elcorto]

Some cleanup / update in network/predictor.py that I'd like to move out of feature branches into develop, in preparation of a larger PR later which includes one feature only.

[RandomDefaultUser]

Hi @elcorto thanks for the improvements, I left a two comments for clarification, apart from that this looks good to me!

[RandomDefaultUser]

I was wondering, should we address the second part of #508 in this PR as well? As you pointed out there, modifying the check should not be difficult.

[elcorto]

I was wondering, should we address the second part of #508 in this PR as well? As you pointed out there, modifying the check should not be difficult.

Ideally yes, but I left it for discussion and did not change anything since I was unsure how to change the check.

[RandomDefaultUser]

I was wondering, should we address the second part of #508 in this PR as well? As you pointed out there, modifying the check should not be difficult.

Ideally yes, but I left it for discussion and did not change anything since I was unsure how to change the check.

Got it, I actually looked into it and do no longer think this part of the code is redundant. The additional check makes sure that if the user specified an inference grid AFTER loading of the predictor class, this change is taken into account. Since the predictor object is loaded and the parameters object is not created beforehand, if you want to use a different inference grid this is the only way (I think?). The part in predict_from_qeout is still redundant, since the check should (and does) happen in predict_from_atoms, so this PR is still good to go from my side, if the CI runs for the docs.
I would close #508 though afterwards, I can add this exact comment to the issue to document it.

[elcorto]

I was wondering, should we address the second part of #508 in this PR as well? As you pointed out there, modifying the check should not be difficult.

Ideally yes, but I left it for discussion and did not change anything since I was unsure how to change the check.

Got it, I actually looked into it and do no longer think this part of the code is redundant. The additional check makes sure that if the user specified an inference grid AFTER loading of the predictor class, this change is taken into account. Since the predictor object is loaded and the parameters object is not created beforehand, if you want to use a different inference grid this is the only way (I think?). The part in predict_from_qeout is still redundant, since the check should (and does) happen in predict_from_atoms, so this PR is still good to go from my side, if the CI runs for the docs. I would close #508 though afterwards, I can add this exact comment to the issue to document it.

OK thanks for the double check. Let's go ahead and close #508 then with this PR.

[RandomDefaultUser]

Sounds good, will merge here.