Skip to content

sklearn/pipeline-fit-params

Rule pipeline-fit-params
Pack scikit-learn
Severity error
Fix none, edit by hand
Library versions every version

Pipeline.fit raises ValueError on a bare sample_weight=, because it takes step parameters as the step name, two underscores and the name. Write it as <step>__sample_weight=.

pipe = make_pipeline(StandardScaler(), LogisticRegression())
pipe.fit(X, y, sample_weight=w)
pipe.fit(X, y, logisticregression__sample_weight=w)

The rule reports pipe.fit(X, y, sample_weight=w) on a Pipeline built in the same block, with metadata routing off.

xpd has no automatic fix, so make the change by hand.

With routing off, Pipeline.fit raises "ValueError: Pipeline.fit does not accept the sample_weight parameter". The rule reads the Pipeline from an assignment earlier in the same block, or from the call itself. A file that turns routing on is left alone, because there a plain name is the right form.

Recall was 23 of 30. The rule found none of 6 sample_weight= keywords given to the fit of a search over a Pipeline, which raises the same ValueError. It also misses a Pipeline built outside a loop and fitted inside it.

xpd prints this message with the issue. A name that starts with $ stands for code from your file.

Pipeline.fit raises ValueError on $K, because it takes step parameters as the step name, two underscores and the name. Write it as <last step>__$K, such as logisticregression__$K for make_pipeline.

Browse the other rules of the scikit-learn pack.