Giordon Stark

So I went ahead and went down the rabbit hole here. I don't know what ROOT is doing, and I had a hard time passing in a TH1 for initial...

One interesting thing is we can always define the `keras_core` layers to inject or patch missing functionality (e.g. shims).

See #2121

I thought we had an issue for this already... @alexander-held we talked about this before...

> @kratsg you have this slated for [`v0.7.1`](https://github.com/scikit-hep/pyhf/projects/19#card-85405165) at the moment. Does this mean that you feel we have a problem that needs to be patched? If not, we should...

Look into how [nbformat](https://github.com/jupyter/nbformat/blob/590cf592d725c09394f9ce5c52db56876c783a18/nbformat/converter.py) handles converting between versions: (thanks @agoose77 ) which is done by defining a `v#/convert.py` such as for `v2` [here](https://github.com/jupyter/nbformat/blob/590cf592d725c09394f9ce5c52db56876c783a18/nbformat/v2/convert.py).

Re-opening due to a user request from [Luis's talk](https://indico.cern.ch/event/1294577/contributions/5677127/) during the 2023 pyhf workshop.

Not sure what you want from the logic. When `hesse` fails, that's because the `minimizer.valid == True` right before calling it, so we'd probably want to just update minuit to...

This will be changed as part of the HS3 migration, with documentation warning under `staterror` similar to what is currently done for `shapesys`. "Can only have one staterror per channel....

I think one issue is that auxdata is required for any modifier that is constrained - and `lumi` is a constrained modifier. We could expand the schema to be more...