[python] [sklearn] respect `eval_at` aliases in keyword arguments (#4599)

python-package/lightgbm/sklearn.py

@@ -26,7 +26,7 @@ class _ObjectiveFunctionWrapper:
         Parameters
         ----------
         func : callable
-            Expects a callable with signature ``func(y_true, y_pred)`` or ``func(y_true, y_pred, group)
+            Expects a callable with signature ``func(y_true, y_pred)`` or ``func(y_true, y_pred, group)``
             and returns (grad, hess):
 
                 y_true : array-like of shape = [n_samples]
@@ -611,9 +611,12 @@ class LGBMModel(_LGBMModelBase):
                 params.pop(alias, None)
             params['num_class'] = self._n_classes
         if hasattr(self, '_eval_at'):
+            eval_at = self._eval_at
             for alias in _ConfigAliases.get('eval_at'):
-                params.pop(alias, None)
-            params['eval_at'] = self._eval_at
+                if alias in params:
+                    _log_warning(f"Found '{alias}' in params. Will use it instead of 'eval_at' argument")
+                    eval_at = params.pop(alias)
+            params['eval_at'] = eval_at
         params['objective'] = self._objective
         if self._fobj:
             params['objective'] = 'None'  # objective = nullptr for unknown objective
@@ -752,7 +755,7 @@ class LGBMModel(_LGBMModelBase):
                 pred_leaf=False, pred_contrib=False, **kwargs):
         """Docstring is set after definition, using a template."""
         if self._n_features is None:
-            raise LGBMNotFittedError("Estimator not fitted, call `fit` before exploiting the model.")
+            raise LGBMNotFittedError("Estimator not fitted, call fit before exploiting the model.")
         if not isinstance(X, (pd_DataFrame, dt_DataTable)):
             X = _LGBMCheckArray(X, accept_sparse=True, force_all_finite=False)
         n_features = X.shape[1]

tests/python_package_test/test_sklearn.py

@@ -143,6 +143,19 @@ def test_xendcg():
     assert gbm.best_score_['valid_0']['ndcg@3'] > 0.6253
 
 
+def test_eval_at_aliases():
+    rank_example_dir = Path(__file__).absolute().parents[2] / 'examples' / 'lambdarank'
+    X_train, y_train = load_svmlight_file(str(rank_example_dir / 'rank.train'))
+    X_test, y_test = load_svmlight_file(str(rank_example_dir / 'rank.test'))
+    q_train = np.loadtxt(str(rank_example_dir / 'rank.train.query'))
+    q_test = np.loadtxt(str(rank_example_dir / 'rank.test.query'))
+    for alias in ('eval_at', 'ndcg_eval_at', 'ndcg_at', 'map_eval_at', 'map_at'):
+        gbm = lgb.LGBMRanker(n_estimators=5, **{alias: [1, 2, 3, 9]})
+        with pytest.warns(UserWarning, match=f"Found '{alias}' in params. Will use it instead of 'eval_at' argument"):
+            gbm.fit(X_train, y_train, group=q_train, eval_set=[(X_test, y_test)], eval_group=[q_test])
+        assert list(gbm.evals_result_['valid_0'].keys()) == ['ndcg@1', 'ndcg@2', 'ndcg@3', 'ndcg@9']
+
+
 def test_regression_with_custom_objective():
     X, y = load_boston(return_X_y=True)
     X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.1, random_state=42)