iminuit integration broken with iminuit==2.0.0

[kosiokarchev]

Hello,
A few days ago the iminuit package was updated to version 2.0.0, and now it seems that the integration with the fit_lc function is broken:

>>> import sncosmo
>>> data = sncosmo.load_example_data()
>>> model = sncosmo.Model(source='SALT2')
>>> sncosmo.fit_lc(data, model, ('t0',))
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/usr/lib/python3.9/site-packages/sncosmo/fitting.py", line 670, in fit_lc
    m = iminuit.Minuit(fitchisq, errordef=1.,
  File "/usr/lib/python3.9/site-packages/iminuit/_minuit.py", line 485, in __init__
    self._init_state = _make_init_state(self._pos2var, args, kwds)
  File "/usr/lib/python3.9/site-packages/iminuit/_minuit.py", line 1401, in _make_init_state
    raise RuntimeError(
RuntimeError: errordef is not one of the parameters []

This works when iminuit is downgraded to 1.5.4 (the latest 1.x version).

PS Thanks for your work on the package, it's a huge relief for someone just getting into the field!

[kboone]

It looks like the entire API was reworked in the latest version of iminuit. The changes are listed here: https://iminuit.readthedocs.io/en/stable/changelog.html

There will need to be some significant changes in sncosmo to get things working with iminuit 2.0. For now, I would recommend sticking with iminuit v1.5.4.

@kbarbary, how do you feel about this? With all of the changes, we'll need to rewrite a good chunk of fit_lc. It looks like that should be pretty straightforward, but it won't be easy to maintain compatibility with both iminuit v1.x and v2.x. We probably need to pick one of the two going forward...

[kboone]

I updated sncosmo to use the new iminuit API in #292, so checkout that branch if you want to use sncosmo with iminuit>=2.0.

[kboone]

#292 has been merged, so this is fixed in the master branch now.