Fix bug in nuts stats

[aseyboldt]

The values for the sampler stats diverging should be a bool, not the whole info tuple. This should fix a problem when using the hdf5 backend (see #2448).

[junpenglao]

while we are on this, I think we should cast effective samples to int as well.

[aseyboldt]

Hm. For some reason there is a test that explicitly checks if it returns a float: https://github.com/pymc-devs/pymc3/blob/master/pymc3/tests/test_diagnostics.py#L192

[junpenglao]

Is there any reason for that? Can we change the test to int as well?

[springcoil]

Is this anything to do with GPU support?

…

On 2 Aug 2017 07:45, "Junpeng Lao" ***@***.***> wrote: Is there any reason for that? Can we change the test to int as well? — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#2467 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AA8DiEgPigbdvZrA2Q_EIATOFtRKYM-pks5sUBrtgaJpZM4OqamT> .

[aseyboldt]

@springcoil No, this was just a bit ugly, and led to problems with the hdf5 backend.
I'm merging this, the only test failure is the good old nuts warning problem.