Closed freddyaboulton closed 2 years ago
I am having the same issue. I cannot downgrade my numpy version due to other packages requiring newer functionality.
I'll look into this. We broke sktime when we tried to do this before (see #423, #425, and https://github.com/alan-turing-institute/sktime/pull/828, and https://github.com/alan-turing-institute/sktime/pull/831 for more detail) and they are much bigger than us, so we pinned it to what it is now. Hopefully I can whip something up!
@aaronreidsmith Thanks for looking into this! I see that #451 is passing all checks but hasn't been merged yet. Do you have an estimate of when it will be merged?
@freddyaboulton I was waiting on @tgsmith61591's approval. It is merged, we'll get a release out with the change soon
Thank you @aaronreidsmith and @tgsmith61591 !
This should be fixed in version 1.8.3 which went out today
Thank you!
Is your feature request related to a problem? Please describe. pmdarima has a pin on numpy 1.19 version which forces uses to use 1.19 across their entire library/application even if they have other dependencies.
This is problematic because if a user's dependency has a pin on numpy>=1.20, then their package is broken.
We saw this in sktime, where numba introduced a lower bound on numpy>=1.20 https://github.com/numba/numba/issues/7339 and we couldn't upgrade numpy because then pmdarima could not be install ed.
Thank you!
Describe the solution you'd like pmdarima depends on numpy>=1.19 or at least allow both numpy 1.19, 1.20
Describe alternatives you've considered
Additional context