def __init__(self, outputs, column_names, column_types, **kwargs):
# TODO: support customized column_types.
inputs = node.StructuredDataInput()
inputs.column_types = column_types
inputs.column_names = column_names
if column_types:
for column_type in column_types.values():
if column_type not in ['categorical', 'numerical']:
raise ValueError(
'Column_types should be either "categorical" '
'or "numerical", but got {name}'.format(
name=column_type))
if column_names and column_types:
for column_name in column_types:
if column_name not in column_names:
raise ValueError(
'Column_names and column_types are '
'mismatched. Cannot find column name '
'{name} in the data.'.format(name=column_name))
super().__init__(inputs=inputs, outputs=outputs, **kwargs)
def set_config(self, config):
self.num_columns = config['num_columns']
self.input_node = node.StructuredDataInput(*config['input_node'])
self.max_columns = config['max_columns']
def test_structured_data_input_transform():
(x, _), _1 = common.dataframe_dataframe()
input_node = node.StructuredDataInput()
input_node.fit(x)
input_node.transform(x)
assert input_node.column_names[0] == 'sex'
def test_structured_data_input_transform():
(x, _), _1 = common.dataframe_dataframe()
input_node = node.StructuredDataInput()
input_node.transform(x)
assert input_node.column_names[0] == 'sex'
assert input_node.column_types == common.COLUMN_TYPES_FROM_CSV
