Fix sklearn compatibility issues (#148)

Author: andersy005

examples/2020ECAHM-scikit-downscale.ipynb

@@ -51,25 +51,34 @@ def get_sample_data(
         return (
             data['uas.hist.CanESM2.CRCM5-UQAM.day.NAM-44i.raw']
             .sel(lat=40.25, lon=-109.2, method='nearest')
-            .to_dataset()
+            .to_dataset()[['uas']]
+            .convert_calendar('standard')
             .squeeze()
             .to_dataframe()[['uas']]
+            .resample('1d')
+            .first()
         )
     elif kind == 'wind-obs':
         return (
             data['uas.gridMET.NAM-44i']
             .sel(lat=40.25, lon=-109.2, method='nearest')
-            .to_dataset()
+            .to_dataset()[['uas']]
+            .convert_calendar('standard')
             .squeeze()
             .to_dataframe()[['uas']]
+            .resample('1d')
+            .first()
         )
     elif kind == 'wind-rcp':
         return (
             data['uas.rcp85.CanESM2.CRCM5-UQAM.day.NAM-44i.raw']
             .sel(lat=40.25, lon=-109.2, method='nearest')
-            .to_dataset()
+            .to_dataset()[['uas']]
+            .convert_calendar('standard')
             .squeeze()
             .to_dataframe()[['uas']]
+            .resample('1d')
+            .first()
         )
     else:
         raise ValueError(kind)

examples/bcsd_example.ipynb

@@ -58,15 +58,18 @@
         "dask>=2024.11",
         "jupyterlab>=4.3",
         "matplotlib>=3.9.4",
+        "mlinsights>=0.5.2",
         "netcdf4>=1.7",
         "numpydoc>=1.9.0",
+        "probscale>=0.2.5",
         "s3fs>=2025.10.0",
         "scipy>=1.14",
+        "seaborn>=0.13.2",
         "sphinx-gallery>=0.19.0",
         "sphinx-rtd-theme>=3.0.2",
         "sphinx>=8.0",
         "zarr>=2.18.2",
-    ]
+]
 
 [tool.ruff]
     builtins = ["ellipsis"]

examples/gard_example.ipynb

@@ -69,7 +69,7 @@ def arrm_breakpoints(X, y, window_width, max_breakpoints):
 
     # lower half of distribution
     # start at 0.4 or the first breakpoint
-    start = min(breakpoints) if breakpoints else start
+    start = min(breakpoints, default=start)
     # likely need this to avoid recomputing r2 and picking the same breakpoint twice
     start -= (min_width // 2) + 1
 
@@ -82,7 +82,7 @@ def arrm_breakpoints(X, y, window_width, max_breakpoints):
         r2[mid] = np.corrcoef(X[s], y[s])[0, 1] ** 2
 
     # find the last three breakpoints
-    for bp in range(max_breakpoints // 2):  # this means max_breakpoints must always be even
+    for _ in range(max_breakpoints // 2):  # this means max_breakpoints must always be even
         mind = np.argmin(r2[:start])  # find minimum r2, only look at the first part of the array
         breakpoints.append(mind)  # breakpoint is in the center of the window
 

examples/utils.py

@@ -29,7 +29,45 @@ def _check_array(self, array, **kwargs):
 
         return array
 
-    def _validate_data(self, X, y=None, reset=True, validate_separately=False, **check_params):
+    def _check_n_features(self, X, reset):
+        """Check and set n_features_in_ attribute.
+
+        Parameters
+        ----------
+        X : array-like
+            Input data
+        reset : bool
+            Whether to reset n_features_in_ or check consistency
+        """
+        n_features = X.shape[1] if hasattr(X, 'shape') and len(X.shape) > 1 else 1
+
+        if reset:
+            self.n_features_in_ = n_features
+        elif hasattr(self, 'n_features_in_'):
+            if self.n_features_in_ != n_features:
+                raise ValueError(
+                    f'X has {n_features} features, but {self.__class__.__name__} '
+                    f'was fitted with {self.n_features_in_} features.'
+                )
+
+    def __sklearn_tags__(self):
+        """Get estimator tags for sklearn 1.6+.
+
+        Returns
+        -------
+        tags : Tags
+            Tags object with estimator metadata.
+        """
+        from dataclasses import replace
+
+        tags = super().__sklearn_tags__()
+        # Update target_tags to indicate y is not required by default
+        tags = replace(tags, target_tags=replace(tags.target_tags, required=False))
+        return tags
+
+    def _validate_data(
+        self, X, y=None, reset: bool = True, validate_separately: bool = False, **check_params: dict
+    ):
         """Validate input data and set or check the `n_features_in_` attribute.
 
         Parameters
@@ -60,7 +98,8 @@ def _validate_data(self, X, y=None, reset=True, validate_separately=False, **che
         """
 
         if y is None:
-            if self._get_tags()['requires_y']:
+            tags = self.__sklearn_tags__()
+            if tags.target_tags.required:
                 raise ValueError(
                     f'This {self.__class__.__name__} estimator '
                     f'requires y to be passed, but the target y is None.'
@@ -80,7 +119,6 @@ def _validate_data(self, X, y=None, reset=True, validate_separately=False, **che
                 X, y = self._check_X_y(X, y, **check_params)
             out = X, y
 
-        # TO-DO: add check_n_features attribute
         if check_params.get('ensure_2d', True):
             self._check_n_features(X, reset=reset)
 

pyproject.toml

@@ -162,10 +162,7 @@ def predict(self, X):
         Xqm = self._qm_transform_by_group(self._create_groups(X, climate_trend=True))
 
         # calculate the anomalies as a ratio of the training data
-        if self.return_anoms:
-            return self._calc_ratio_anoms(Xqm, self.y_climo_)
-        else:
-            return Xqm
+        return self._calc_ratio_anoms(Xqm, self.y_climo_) if self.return_anoms else Xqm
 
     def _calc_ratio_anoms(self, obj, climatology, climate_trend=False):
         """helper function for dividing day groups by climatology"""
@@ -181,29 +178,13 @@ def _calc_ratio_anoms(self, obj, climatology, climate_trend=False):
 
         return result
 
-    def _more_tags(self):
-        return {
-            '_xfail_checks': {
-                'check_estimators_dtypes': 'BCSD only suppers 1 feature',
-                'check_dtype_object': 'BCSD only suppers 1 feature',
-                'check_fit_score_takes_y': 'BCSD only suppers 1 feature',
-                'check_estimators_fit_returns_self': 'BCSD only suppers 1 feature',
-                'check_estimators_fit_returns_self(readonly_memmap=True)': 'BCSD only suppers 1 feature',
-                'check_pipeline_consistency': 'BCSD only suppers 1 feature',
-                'check_estimators_nan_inf': 'BCSD only suppers 1 feature',
-                'check_estimators_overwrite_params': 'BCSD only suppers 1 feature',
-                'check_estimators_pickle': 'BCSD only suppers 1 feature',
-                'check_fit2d_predict1d': 'BCSD only suppers 1 feature',
-                'check_methods_subset_invariance': 'BCSD only suppers 1 feature',
-                'check_fit2d_1sample': 'BCSD only suppers 1 feature',
-                'check_dict_unchanged': 'BCSD only suppers 1 feature',
-                'check_dont_overwrite_parameters': 'BCSD only suppers 1 feature',
-                'check_fit_idempotent': 'BCSD only suppers 1 feature',
-                'check_n_features_in': 'BCSD only suppers 1 feature',
-                'check_fit_check_is_fitted': 'BCSD only suppers 1 feature',
-                'check_methods_sample_order_invariance': 'temporal order matters',
-            },
-        }
+    def __sklearn_tags__(self):
+        from dataclasses import replace
+
+        tags = super().__sklearn_tags__()
+        # Skip tests - only supports 1 feature, temporal order matters
+        tags = replace(tags, _skip_test='BCSD only supports 1 feature and temporal order matters')
+        return tags
 
 
 class BcsdTemperature(BcsdBase):
@@ -283,38 +264,20 @@ def rolling_func(x):
 
     def _remove_climatology(self, obj, climatology, climate_trend=False):
         """helper function to remove climatologies"""
-        dfs = []
-        for key, group in self._create_groups(obj, climate_trend):
-            if self.timestep == 'monthly':
-                dfs.append(group - climatology.loc[key].values)
-            elif self.timestep == 'daily':
-                dfs.append(group - climatology.loc[key].values)
-
+        dfs = [
+            group - climatology.loc[key].values
+            for key, group in self._create_groups(obj, climate_trend)
+            if self.timestep in ['monthly', 'daily']
+        ]
         result = pd.concat(dfs).sort_index()
         if obj.shape != result.shape:
             raise ValueError('shape of climo is not equal to input array')
         return result
 
-    def _more_tags(self):
-        return {
-            '_xfail_checks': {
-                'check_estimators_dtypes': 'BCSD only suppers 1 feature',
-                'check_fit_score_takes_y': 'BCSD only suppers 1 feature',
-                'check_estimators_fit_returns_self': 'BCSD only suppers 1 feature',
-                'check_estimators_fit_returns_self(readonly_memmap=True)': 'BCSD only suppers 1 feature',
-                'check_dtype_object': 'BCSD only suppers 1 feature',
-                'check_pipeline_consistency': 'BCSD only suppers 1 feature',
-                'check_estimators_nan_inf': 'BCSD only suppers 1 feature',
-                'check_estimators_overwrite_params': 'BCSD only suppers 1 feature',
-                'check_estimators_pickle': 'BCSD only suppers 1 feature',
-                'check_fit2d_predict1d': 'BCSD only suppers 1 feature',
-                'check_methods_subset_invariance': 'BCSD only suppers 1 feature',
-                'check_fit2d_1sample': 'BCSD only suppers 1 feature',
-                'check_dict_unchanged': 'BCSD only suppers 1 feature',
-                'check_dont_overwrite_parameters': 'BCSD only suppers 1 feature',
-                'check_fit_idempotent': 'BCSD only suppers 1 feature',
-                'check_n_features_in': 'BCSD only suppers 1 feature',
-                'check_fit_check_is_fitted': 'BCSD only suppers 1 feature',
-                'check_methods_sample_order_invariance': 'temporal order matters',
-            },
-        }
+    def __sklearn_tags__(self):
+        from dataclasses import replace
+
+        tags = super().__sklearn_tags__()
+        # Skip tests - only supports 1 feature, temporal order matters
+        tags = replace(tags, _skip_test='BCSD only supports 1 feature and temporal order matters')
+        return tags