Tests for dask skl categorical data support. (#7054)

python-package/xgboost/sklearn.py

@@ -632,6 +632,12 @@ def _configure_fit(
                 eval_metric = None
             else:
                 params.update({"eval_metric": eval_metric})
+        if self.enable_categorical and params.get("tree_method", None) != "gpu_hist":
+            raise ValueError(
+                "Experimental support for categorical data is not implemented for"
+                " current tree method yet."
+            )
+
         return model, feval, params
 
     def _set_evaluation_result(self, evals_result: TrainingCallback.EvalsLog) -> None:

tests/python-gpu/test_gpu_with_dask.py

@@ -211,20 +211,34 @@ def test_categorical(local_cuda_cluster: LocalCUDACluster) -> None:
         )
         assert tm.non_increasing(by_builtin_results["Train"]["rmse"])
 
-        model = output["booster"]
-        with tempfile.TemporaryDirectory() as tempdir:
-            path = os.path.join(tempdir, "model.json")
-            model.save_model(path)
-            with open(path, "r") as fd:
-                categorical = json.load(fd)
-
-            categories_sizes = np.array(
-                categorical["learner"]["gradient_booster"]["model"]["trees"][-1][
-                    "categories_sizes"
-                ]
-            )
-            assert categories_sizes.shape[0] != 0
-            np.testing.assert_allclose(categories_sizes, 1)
+        def check_model_output(model: dxgb.Booster) -> None:
+            with tempfile.TemporaryDirectory() as tempdir:
+                path = os.path.join(tempdir, "model.json")
+                model.save_model(path)
+                with open(path, "r") as fd:
+                    categorical = json.load(fd)
+
+                categories_sizes = np.array(
+                    categorical["learner"]["gradient_booster"]["model"]["trees"][-1][
+                        "categories_sizes"
+                    ]
+                )
+                assert categories_sizes.shape[0] != 0
+                np.testing.assert_allclose(categories_sizes, 1)
+
+        check_model_output(output["booster"])
+        reg = dxgb.DaskXGBRegressor(
+            enable_categorical=True, n_estimators=10, tree_method="gpu_hist"
+        )
+        reg.fit(X, y)
+
+        check_model_output(reg.get_booster())
+
+        reg = dxgb.DaskXGBRegressor(
+            enable_categorical=True, n_estimators=10
+        )
+        with pytest.raises(ValueError):
+            reg.fit(X, y)
 
 
 def to_cp(x: Any, DMatrixT: Type) -> Any: