Making a 0.9 release, compatible with scikit-learn>=0.24, and one with the new features would be nice.

0.9.0 release (Already released) (bugfixes, compatibility with scikit-learn>=0.24):

• Merge Remove BayesSearchCV(iid=) parameter deprecated in sklearn 0.24 #988 (fix many issues of scikit-learn 0.24 compat)
• Merge [MRG] Fix GaussianProcessRegressor for scikit-learn 1.0 #1063
• Merge [MRG] update packges in circleci deps #1069
• DOC: Update whats_new for 0.9.0 #1075

Unmerged PR:

• ENH: Add constrained optimization #971
• GPR with noise fix #1012
• [WIP] Evaluate random point instead of using point twice #1050
• [MRG] Keep order of variables in LabelEncoder #1055
• ENH: Add support for Multimetric BayesSearchCV #1062
• MNT: fix plot_evaluation when skipping dimension, add test #1066 (fix plot_evaluations() incompatible with BayesSearchCV #1056 )
• MNT: improved check_dimension type inference, added warnings, tests #1067 (fix integer lower_bound and upper_bound result in ignoring "prior", and bug if are not of the same type #1065)

Unofficial test branch
I have made an unofficial test branch on my fork, unofficial, for testing some of those features. If someone wishes to test features & fixes, one can install from source using

pip install git+https://github.com/QuentinSoubeyran/scikit-optimize.git@unofficial

Warning: If you are using this unofficial branch, before reporting an issue:

• If the issue is reproducible with the master branch from this repo, open an issue in this repo
• If not, identify which PR causes the issue, and add a comment in the discussion for this PR

See also:
#1059

EDIT: edited plan to reflect @xmatthias comment
EDIT: edited to reflect release of 0.9.0 version
EDIT: update test branch, formatting