Closed gbordyugov closed 7 years ago
hmm, it fails with minor diffs in floating point constants in low-level bayesian functions that I didn't touch...
I would say that the difference is negligible, from my point of view you can simply update the unit test values.
I'm running tox
-based tests locally in order to reproduce (just used runtests
before)
Thank you Grisha. It looks good! See my comments above for a few small suggestions.
OCTO-1502
introducing
make_delta()
to hide statistics parameters in the call chains.