It does not seem to work as expected:

https://182965-843222-gh.circle-artifacts.com/0/doc/modules/compose.html#visualizing-composite-estimators

Actually it is, you're missing the # doctest: +SKIP :)

Actually it is, you're missing the # doctest: +SKIP :)

Indeed. I forgot that the code of code snippets in the user guide is not executed by sphinx as done in the examples... We already have the problem on the dev site:

https://scikit-learn.org/dev/modules/compose.html#visualizing-composite-estimators

so this is not a problem of this PR.

Hi all,
just wanted to share an experience with the API change. It was solved after I included the following language at the beginning of the notebook, but I think it would be useful for others to be aware of error messages that arise as a result of this change (I believe).

After I updated from v1.0.2 to v1.1.2, a normal notebook calling a sklearn.ensemble.RandomForestRegression object would yield the following error in this actions result.

I then included the language as follows at the beginning of notebooks, and it did improve.

from sklearn import set_config
set_config(display='text')

So long story short is that this change in the default options might introduce some hard-to-debut errors, such as in this case a TypeError.

The issue comes from RegrsesionBenchmark.get_params returning estimator classes as values. I opened #24512 to support this use case. For RegrsesionBenchmark, it will look like this:

Many thanks for this fast iteration - it will be very welcome indeed if this use case is supported! Not only it would align with the other use cases, but it also would take away the need to re-adjust the parameter to its older default.