[pyspark] Remove the deprecated use_gpu parameter. (#11554)

Author: trivialfis

python-package/xgboost/spark/core.py

@@ -123,7 +123,6 @@
     "qid_col",
     "repartition_random_shuffle",
     "pred_contrib_col",
-    "use_gpu",
     "launch_tracker_on_driver",
     "coll_cfg",
 ]
@@ -218,16 +217,6 @@ class _SparkXGBParams(
         ),
         TypeConverters.toString,
     )
-    use_gpu = Param(
-        Params._dummy(),
-        "use_gpu",
-        (
-            "Deprecated, use `device` instead. A boolean variable. Set use_gpu=true "
-            "if the executors are running on GPU instances. Currently, only one GPU per"
-            " task is supported."
-        ),
-        TypeConverters.toBoolean,
-    )
     force_repartition = Param(
         Params._dummy(),
         "force_repartition",
@@ -503,9 +492,7 @@ def _validate_params(self) -> None:
     def _run_on_gpu(self) -> bool:
         """If train or transform on the gpu according to the parameters"""
 
-        return use_cuda(self.getOrDefault(self.device)) or self.getOrDefault(
-            self.use_gpu
-        )
+        return use_cuda(self.getOrDefault(self.device))
 
     def _col_is_defined_not_empty(self, param: "Param[str]") -> bool:
         return self.isDefined(param) and self.getOrDefault(param) not in (None, "")
@@ -628,7 +615,6 @@ def __init__(self) -> None:
         self._setDefault(
             num_workers=1,
             device="cpu",
-            use_gpu=False,
             force_repartition=False,
             repartition_random_shuffle=False,
             feature_names=None,

python-package/xgboost/spark/estimator.py

@@ -1,9 +1,8 @@
 """Xgboost pyspark integration submodule for estimator API."""
 
-# pylint: disable=fixme, protected-access, no-member, invalid-name
+# pylint: disable=protected-access, no-member, invalid-name
 # pylint: disable=unused-argument, too-many-locals
 
-import warnings
 from typing import Any, List, Optional, Type, Union
 
 import numpy as np
@@ -77,12 +76,6 @@ def set_param_attrs(attr_name: str, param: Param) -> None:
         set_param_attrs(name, param_obj)
 
 
-def _deprecated_use_gpu() -> None:
-    warnings.warn(
-        "`use_gpu` is deprecated since 2.0.0, use `device` instead", FutureWarning
-    )
-
-
 class SparkXGBRegressor(_SparkXGBEstimator):
     """SparkXGBRegressor is a PySpark ML estimator. It implements the XGBoost regression
     algorithm based on XGBoost python library, and it can be used in PySpark Pipeline
@@ -140,11 +133,6 @@ class SparkXGBRegressor(_SparkXGBEstimator):
     num_workers:
         How many XGBoost workers to be used to train.
         Each XGBoost worker corresponds to one spark task.
-    use_gpu:
-        .. deprecated:: 2.0.0
-
-        Use `device` instead.
-
     device:
 
         .. versionadded:: 2.0.0
@@ -214,7 +202,6 @@ def __init__(  # pylint:disable=too-many-arguments
         weight_col: Optional[str] = None,
         base_margin_col: Optional[str] = None,
         num_workers: int = 1,
-        use_gpu: Optional[bool] = None,
         device: Optional[str] = None,
         force_repartition: bool = False,
         repartition_random_shuffle: bool = False,
@@ -225,8 +212,6 @@ def __init__(  # pylint:disable=too-many-arguments
     ) -> None:
         super().__init__()
         input_kwargs = self._input_kwargs
-        if use_gpu:
-            _deprecated_use_gpu()
         self.setParams(**input_kwargs)
 
     @classmethod
@@ -327,11 +312,6 @@ class SparkXGBClassifier(_SparkXGBEstimator, HasProbabilityCol, HasRawPrediction
     num_workers:
         How many XGBoost workers to be used to train.
         Each XGBoost worker corresponds to one spark task.
-    use_gpu:
-        .. deprecated:: 2.0.0
-
-        Use `device` instead.
-
     device:
 
         .. versionadded:: 2.0.0
@@ -401,7 +381,6 @@ def __init__(  # pylint:disable=too-many-arguments
         weight_col: Optional[str] = None,
         base_margin_col: Optional[str] = None,
         num_workers: int = 1,
-        use_gpu: Optional[bool] = None,
         device: Optional[str] = None,
         force_repartition: bool = False,
         repartition_random_shuffle: bool = False,
@@ -416,8 +395,6 @@ def __init__(  # pylint:disable=too-many-arguments
         # binary or multinomial input dataset, and we need to remove the fixed default
         # param value as well to avoid causing ambiguity.
         input_kwargs = self._input_kwargs
-        if use_gpu:
-            _deprecated_use_gpu()
         self.setParams(**input_kwargs)
         self._setDefault(objective=None)
 
@@ -517,11 +494,6 @@ class SparkXGBRanker(_SparkXGBEstimator):
     num_workers:
         How many XGBoost workers to be used to train.
         Each XGBoost worker corresponds to one spark task.
-    use_gpu:
-        .. deprecated:: 2.0.0
-
-        Use `device` instead.
-
     device:
 
         .. versionadded:: 2.0.0
@@ -597,7 +569,6 @@ def __init__(  # pylint:disable=too-many-arguments
         base_margin_col: Optional[str] = None,
         qid_col: Optional[str] = None,
         num_workers: int = 1,
-        use_gpu: Optional[bool] = None,
         device: Optional[str] = None,
         force_repartition: bool = False,
         repartition_random_shuffle: bool = False,
@@ -608,8 +579,6 @@ def __init__(  # pylint:disable=too-many-arguments
     ) -> None:
         super().__init__()
         input_kwargs = self._input_kwargs
-        if use_gpu:
-            _deprecated_use_gpu()
         self.setParams(**input_kwargs)
 
     @classmethod

python-package/xgboost/spark/params.py

@@ -39,7 +39,7 @@ class HasBaseMarginCol(Params):
 class HasFeaturesCols(Params):
     """
     Mixin for param features_cols: a list of feature column names.
-    This parameter is taken effect only when use_gpu is enabled.
+    This parameter is taken effect only when GPU is enabled.
     """
 
     features_cols = Param(

tests/test_distributed/test_gpu_with_spark/test_gpu_spark.py

@@ -208,7 +208,7 @@ def test_cv_sparkxgb_classifier_feature_cols_with_gpu(spark_iris_dataset_feature
     assert f1 >= 0.97
 
     clf = SparkXGBClassifier(
-        features_col=feature_names, use_gpu=True, num_workers=num_workers
+        features_col=feature_names, device="cuda", num_workers=num_workers
     )
     grid = ParamGridBuilder().addGrid(clf.max_depth, [6, 8]).build()
     evaluator = MulticlassClassificationEvaluator(metricName="f1")