Reloading issues with older ensemble models (from v1.3) with new sklearn version 1.4

[ARY2260]

I was trying to run .predict() on RandomForestRegressor model trained using v1.3 and then reloaded with v1.4, when I got this error:

AttributeError: 'DecisionTreeRegressor' object has no attribute 'monotonic_cst'

I noticed a new feature update in v1.4 change logs:

Feature ensemble.RandomForestClassifier, ensemble.RandomForestRegressor, ensemble.ExtraTreesClassifier and ensemble.ExtraTreesRegressor now support monotonic constraints, useful when features are supposed to have a positive/negative effect on the target. Missing values in the train data and multi-output targets are not supported. #13649 by Samuel Ronsin, initiated by Patrick O’Reilly.

Does anyone know how to load older models in the v1.4 version of sklearn?

[glemaitre]

We don't support between version unpickling: https://scikit-learn.org/stable/model_persistence.html

Here the reason is that the monotonic constraint has been implement in 1.4 and is therefore not available in 1.3. The internal code has changed between the two versions making the older model not compatible with 1.4