Nathan Simpson

@kratsg should I wait on your suggested refactor of the above? or do you think we could hack in a short-term solution?

@kratsg I looked at the changes, but the only actual code change looked to be the removal of `self.__staterror_uncrt` -- i've left it in, but can you confirm if that...

> > @kratsg I looked at the changes, but the only actual code change looked to be the removal of `self.__staterror_uncrt` -- i've left it in, but can you confirm...

cc'ing @kratsg about this again (and the relevant PR in #742)

Great to see efforts on this again, super exciting for me in particular :) ~Just as an FYI, if you're able to commit to your user forks of pyhf for...

> @kratsg @phinate can you comment a bit more on the specifics of what is causing this problem and/or create a public workspace that will fail for this? > >...

For archival purposes: here's an example of a model with a histosys being succsessfully hacked to mimic another model :) ```python3 import pyhf from math import prod import numpy as...

@kratsg A mwe of this failing: ``` import pyhf pyhf.set_backend('jax') def from_spec(yields): s, b, bup, bdown = yields spec = { "channels": [ { "name": "nn", "samples": [ { "name":...

Thanks for the fix, that works great! Now, we're dealing with the same errors I encountered when I forked this originally: `default_backend` being numpy, even when overriden by `pyhf.default_backend =...

Awesome!! Thanks guys (and @jakevdp), that was the quickest turnaround for an external bug I’ve ever seen...