Fix usage of optuna distributions

[fraimondo]

Using something like this:

creator.add(
    "rf",
    C=IntDistribution(low=100, high=400, step=50),
    max_depth=5,
    min_samples_split=2,
    min_samples_leaf=2,
)

Would not work on scikit-learn 1.3, raising the errror:

sklearn.utils._param_validation.InvalidParameterError: The 'n_estimators' parameter of RandomForestClassifier must be an int in the range [1, inf). Got IntDistribution(high=400, log=False, low=100, step=50) instead.

Apparently this is not an issue in the new scikit-learn (or even in the test suit).

This PR fixes that.

[codecov]

Codecov Report

Attention: Patch coverage is 50.00000% with 3 lines in your changes missing coverage. Please review.

Project coverage is 90.03%. Comparing base (c93ef00) to head (d501297).

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #268      +/-   ##
==========================================
- Coverage   90.13%   90.03%   -0.10%     
==========================================
  Files          54       54              
  Lines        2423     2429       +6     
  Branches      491      492       +1     
==========================================
+ Hits         2184     2187       +3     
- Misses        157      159       +2     
- Partials       82       83       +1     

Flag	Coverage Δ
docs	100.00% <ø> (ø)
julearn	90.03% <50.00%> (-0.10%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files	Coverage Δ
julearn/model_selection/_optuna_searcher.py	83.72% <100.00%> (+1.22%)	⬆️
julearn/pipeline/pipeline_creator.py	86.47% <0.00%> (-0.77%)	⬇️

[github-actions]

PR Preview Action v1.4.7
Preview removed because the pull request was closed.
2024-06-19 16:03 UTC