there is assertion error in case test_crossvalidator of tests/test_tuning.py:99

[johnnyzhon]

Environment:
4 V100 gpu
spark-rapids-ml branch-23.08
CUML 23.08
CUDA 12.1
python3.9
spark3.4.0

pytest error:
src/spark_rapids_ml/tree.py:260: ValueError
_________________ test_crossvalidator[(100, 8)-float32-vector] _________________

tmp_path = '/tmp/spark_rapids_ml_tests_3ywlntku', feature_type = 'vector'
data_type = <class 'numpy.float32'>, data_shape = (100, 8)

@pytest.mark.parametrize("feature_type", [feature_types.vector])
@pytest.mark.parametrize("data_type", [np.float32])
@pytest.mark.parametrize("data_shape", [(100, 8)], ids=idfn)
def test_crossvalidator(
    tmp_path: str,
    feature_type: str,
    data_type: np.dtype,
    data_shape: Tuple[int, int],
) -> None:
    X, _, y, _ = make_regression_dataset(
        datatype=data_type,
        nrows=data_shape[0],
        ncols=data_shape[1],
    )

    with CleanSparkSession() as spark:
        df, features_col, label_col = create_pyspark_dataframe(
            spark, feature_type, data_type, X, y
        )
        assert label_col is not None

        rfc = RandomForestRegressor()
        rfc.setFeaturesCol(features_col)
        rfc.setLabelCol(label_col)

        evaluator = RegressionEvaluator()
        evaluator.setLabelCol(label_col)

        grid = ParamGridBuilder().addGrid(rfc.maxBins, [3, 5]).build()

        cv = CrossValidator(
            estimator=rfc,
            estimatorParamMaps=grid,
            evaluator=evaluator,
            numFolds=2,
            seed=101,
        )

        def check_cv(cv_est: Union[CrossValidator, CrossValidatorModel]) -> None:
            assert isinstance(cv_est, (CrossValidator, CrossValidatorModel))
            assert isinstance(cv_est.getEstimator(), RandomForestRegressor)
            assert isinstance(cv_est.getEvaluator(), RegressionEvaluator)
            assert cv_est.getNumFolds() == 2
            assert cv_est.getSeed() == 101
            assert cv_est.getEstimatorParamMaps() == grid

        check_cv(cv)

        path = tmp_path + "/cv"
        cv_path = f"{path}/cv"

        cv.write().overwrite().save(cv_path)
        cv_loaded = CrossValidator.load(cv_path)

        check_cv(cv_loaded)

        cv_model = cv.fit(df)
        check_cv(cv_model)

        cv_model_path = f"{path}/cv-model"
        cv_model.write().overwrite().save(cv_model_path)
        cv_model_loaded = CrossValidatorModel.load(cv_model_path)

        check_cv(cv_model_loaded)

      assert evaluator.evaluate(cv_model.transform(df)) == evaluator.evaluate(

            cv_model_loaded.transform(df)
        )

E assert 129.9102859946124 == 129.91028599461237
E + where 129.9102859946124 = (DataFrame[label_col: float, features: array, prediction: double])
E + where = RegressionEvaluator_c12c2d118a71.evaluate
E + and DataFrame[label_col: float, features: array, prediction: double] = (DataFrame[label_col: float, features: vector])
E + where = CrossValidatorModel_4072bbc9e0e1.transform
E + and 129.91028599461237 = (DataFrame[label_col: float, features: array, prediction: double])
E + where = RegressionEvaluator_c12c2d118a71.evaluate
E + and DataFrame[label_col: float, features: array, prediction: double] = (DataFrame[label_col: float, features: vector])
E + where = CrossValidatorModel_4072bbc9e0e1.transform

tests/test_tuning.py:99: AssertionError