update

Signed-off-by: Weichen Xu <weichen.xu@databricks.com>

python-package/xgboost/compat.py

@@ -76,13 +76,13 @@ def lazy_isinstance(instance: Any, module: str, name: str) -> bool:
 _logger = logging.getLogger(__name__)
 
 
-def is_cudf_installed() -> bool:
-    """Check cuDF installed or not"""
-    # Checking by `importing` instead of check `importlib.util.find_spec("cudf") is not None`
-    # because user might install cudf successfully but importing cudf raises issues (e.g. saying
-    # running on mismatched cuda version)
+def is_cudf_available() -> bool:
+    """Check cuDF package available or not"""
+    if importlib.util.find_spec("cudf") is None:
+        return False
     try:
         import cudf
+
         return True
     except ImportError:
         _logger.exception("Importing cuDF failed, use DMatrix instead of QDM")

python-package/xgboost/spark/core.py

@@ -32,12 +32,12 @@
     ShortType,
 )
 from scipy.special import expit, softmax  # pylint: disable=no-name-in-module
+from xgboost.compat import is_cudf_available
 from xgboost.core import Booster
 from xgboost.training import train as worker_train
 
 import xgboost
 from xgboost import XGBClassifier, XGBRanker, XGBRegressor
-from xgboost.compat import is_cudf_installed
 
 from .data import (
     _read_csr_matrix_from_unwrapped_spark_vec,
@@ -773,10 +773,10 @@ def _train_booster(pandas_df_iter):
 
             # If cuDF is not installed, then using DMatrix instead of QDM,
             # because without cuDF, DMatrix performs better than QDM.
-            # Note: Checking `is_cudf_installed` in spark worker side because
+            # Note: Checking `is_cudf_available` in spark worker side because
             # spark worker might has different python environment with driver side.
             if use_gpu:
-                use_qdm = use_hist and is_cudf_installed()
+                use_qdm = use_hist and is_cudf_available()
             else:
                 use_qdm = use_hist