How to use the matminer.featurizers.composition.ElementProperty function in matminer

def all(self):
        fs = [
            cf.AtomicOrbitals(),
            cf.ElementProperty.from_preset("matminer"),
            cf.ElementProperty.from_preset("magpie"),
            cf.ElementProperty.from_preset("matscholar_el"),
            cf.ElementProperty.from_preset("deml"),
            cf.Meredig(),
            cf.ElementFraction(),
            cf.Stoichiometry(),
            cf.TMetalFraction(),
            cf.BandCenter(),
            cf.ValenceOrbital(),
            cf.YangSolidSolution(),
            cf.CationProperty.from_preset(preset_name="deml"),
            cf.OxidationStates.from_preset(preset_name="deml"),
            cf.ElectronAffinity(),
            cf.ElectronegativityDiff(),
            cf.IonProperty(fast=True),
            cf.Miedema(),
            cf.AtomicPackingEfficiency(),  # slower than the rest
            cf.CohesiveEnergy(),  # requires mpid present
        ]

def X_train(self):
        return self._X_train

    @property
    def y_train(self):
        return self._y_train

if __name__ == '__main__':
    from automatminer.data.load import load_glass_formation
    from pymatgen.core import Composition
    from matminer.featurizers.composition import ElementProperty

    df = load_glass_formation()
    df['composition'] = df["formula"].apply(lambda x: Composition(x))

    elemprop = ElementProperty.from_preset("matminer")
    df = elemprop.featurize_dataframe(df, col_id="composition")

    feature_cols = elemprop.feature_labels()
    target = "gfa"

    automl = AutoSklearnML(X=df[feature_cols],
                           y=df[target],
                           dataset_name="ternary glass formation",
                           time_left_for_this_task=60,
                           per_run_time_limit=30,
                           output_folder="/tmp/matbench_automl/tmp",
                           tmp_folder="/tmp/matbench_automl/out")
    automl.classification()

pd.set_option('display.height', 1000)
pd.set_option('display.max_rows', 500)
pd.set_option('display.max_columns', 500)
pd.set_option('display.width', 1000)

# Try predict ehull from initial structure
n = 500
print("Reading csv for {} compounds...".format(n))
df = load_mp('mp_all.csv').sample(n=n)
print("Constructing {} structures from dictionaries...".format(n))
df['structure'] = [Structure.from_dict(s) for s in df['structure']]
df['initial structure'] = [Structure.from_dict(s) for s in df['initial structure']]
df['composition'] = [f.composition for f in df['structure']]

# Pick featurizers
ep = ElementProperty.from_preset("matminer")
bb = BagofBonds()
bf = BondFractions(approx_bonds=True)
sh = StructuralHeterogeneity()
co = ChemicalOrdering()
de = DensityFeatures()
composition_featurizers = [ep]
structure_featurizers = [bf, sh, co, de]


# Featurizing
fls = []
for cf in composition_featurizers:
    print("Featurizing {}...".format(cf.__class__.__name__))
    cf.fit_featurize_dataframe(df, 'composition', ignore_errors=True)
    fls += cf.feature_labels()