MNT Remove deprecated iprint and disp usage in scipy 1.15 LBFGS
#31642
Conversation
lesteve
changed the title
iprint and disp usage in scipy 1.15 LBFGS
Co-authored-by: Tim Head <betatim@gmail.com>
It feels like the main test is that there is no warning (turned into error) into the scipy-dev build (or with scipy 1.16 when we update the lock-file and we actually use scipy 1.16)? |
betatim
added
Waiting for Second Reviewer
|
The error (emission of DeprecationWarning) fixed by this PR can be seen in tests of the debian package against scipy 1.16.0, https://ci.debian.net/data/autopkgtest/unstable/amd64/s/scikit-learn/61828718/log.gz Note there seems to be a separate scipy 1.16.0 error in test_logistic_regression_path_convergence_fail. |
There was a problem hiding this comment.
LGTM. Thanks @lesteve
Let's merge as is to free the CI. In the future we may have to update or possibly even deprecate the verbose parameter of some estimators since it's usually what controls iprint
Fix #31624
iprintanddispdo not do anything since scipy 1.15 and for scipy >= 1.16 aDeprecationWarningis emitted, see scipy/scipy#23186 (comment).I guess a side-effect is that for people who cared about this additional info, it won't be available any more with scipy >= 1.15 but this seems like a scipy decision and I am not sure we can do much about it, unless maybe there is another way to have a similar info?