Fix GPU L1 error. (#8749)

tests/cpp/tree/test_node_partition.cc

@@ -0,0 +1,24 @@
+/**
+ * Copyright 2023 by XGBoost contributors
+ */
+#include <gtest/gtest.h>
+#include <xgboost/task.h>
+#include <xgboost/tree_updater.h>
+
+namespace xgboost {
+TEST(Updater, HasNodePosition) {
+  Context ctx;
+  ObjInfo task{ObjInfo::kRegression, true, true};
+  std::unique_ptr<TreeUpdater> up{TreeUpdater::Create("grow_histmaker", &ctx, task)};
+  ASSERT_TRUE(up->HasNodePosition());
+
+  up.reset(TreeUpdater::Create("grow_quantile_histmaker", &ctx, task));
+  ASSERT_TRUE(up->HasNodePosition());
+
+#if defined(XGBOOST_USE_CUDA)
+  ctx.gpu_id = 0;
+  up.reset(TreeUpdater::Create("grow_gpu_hist", &ctx, task));
+  ASSERT_TRUE(up->HasNodePosition());
+#endif  // defined(XGBOOST_USE_CUDA)
+}
+}  // namespace xgboost

tests/python-gpu/test_gpu_prediction.py

@@ -337,13 +337,21 @@ class TestGPUPredict:
     @given(predict_parameter_strategy, tm.dataset_strategy)
     @settings(deadline=None, max_examples=20, print_blob=True)
     def test_predict_leaf_gbtree(self, param, dataset):
+        # Unsupported for random forest
+        if param.get("num_parallel_tree", 1) > 1 and dataset.name.endswith("-l1"):
+            return
+
         param['booster'] = 'gbtree'
         param['tree_method'] = 'gpu_hist'
         self.run_predict_leaf_booster(param, 10, dataset)
 
     @given(predict_parameter_strategy, tm.dataset_strategy)
     @settings(deadline=None, max_examples=20, print_blob=True)
-    def test_predict_leaf_dart(self, param, dataset):
+    def test_predict_leaf_dart(self, param: dict, dataset: tm.TestDataset) -> None:
+        # Unsupported for random forest
+        if param.get("num_parallel_tree", 1) > 1 and dataset.name.endswith("-l1"):
+            return
+
         param['booster'] = 'dart'
         param['tree_method'] = 'gpu_hist'
         self.run_predict_leaf_booster(param, 10, dataset)

tests/python/test_updaters.py

@@ -442,6 +442,22 @@ class TestTreeMethod:
         config_0 = json.loads(booster_0.save_config())
         np.testing.assert_allclose(get_score(config_0), get_score(config_1) + 1)
 
+        evals_result: Dict[str, Dict[str, list]] = {}
+        xgb.train(
+            {
+                "tree_method": tree_method,
+                "objective": "reg:absoluteerror",
+                "subsample": 0.8
+            },
+            Xy,
+            num_boost_round=10,
+            evals=[(Xy, "Train")],
+            evals_result=evals_result,
+        )
+        mae = evals_result["Train"]["mae"]
+        assert mae[-1] < 20.0
+        assert tm.non_increasing(mae)
+
     @pytest.mark.skipif(**tm.no_sklearn())
     @pytest.mark.parametrize(
         "tree_method,weighted", [

tests/test_distributed/test_with_spark/test_spark_local.py

@@ -215,7 +215,6 @@ MultiClfData = namedtuple("MultiClfData", ("multi_clf_df_train", "multi_clf_df_t
 
 @pytest.fixture
 def multi_clf_data(spark: SparkSession) -> Generator[MultiClfData, None, None]:
-
     X = np.array([[1.0, 2.0, 3.0], [1.0, 2.0, 4.0], [0.0, 1.0, 5.5], [-1.0, -2.0, 1.0]])
     y = np.array([0, 0, 1, 2])
     cls1 = xgb.XGBClassifier()