Giordon Stark

> Using `test_stat="q"` indeed solved the NaN problem, however changing the lower bound to a negative number doesn't seem to be necessary but it slightly change the returned CLs. Thanks...

For what it's worth, I took the existing code and ran it through `minuit` by changing the top line of the code to set the minuit optimizer in the backend:...

For some more progress, seems like some of the normfactors do hit the parameters at the limit during the non-gradient version of the optimization (`do_grad=False`). This corresponds to one of...

> As it was the first analysis to have several fixed parameters, and all previous analyses were working great, maybe it could be related to that? likely not. fixed parameters...

@sabinekraml @Ga0l -- I've checked this against our latest release (`0.6.0`) and I get the following ``` $ python test.py # mu = 0.9999 (tensor(0.9869), tensor(0.9868)) $ python test.py #...

If everyone looks ok, please feel free to close the issue. Specifically, scipy/scipy#13009 is one of the big improvements we see in 1.6.0 of scipy.

> For any code that builds many `Model`s often (as some of mine does), this 50x slowdown can suddenly dominate the execution time. The network request should at least be...

> I expect schema validation to succeed without crashing even when there is no network access that allows resolving the https schema-ids. Yeah, this is a regression. I thought I...

> You can see the first [`Timer.autorange()` call](https://github.com/python/cpython/blob/1bc13df6281e72c6a5bbc5b2b6b17b95b829d14b/Lib/timeit.py#L319-L326) takes 0.491 s, which is with a cold cache. The rest of the calls take ~0.2 s, which is basically just the...

Interesting, then `store` seems pretty useless on its own since I would have assumed it would pick through that first based on documentation > A mapping from URIs to documents...