def _meta_build(self, dataset):
# Using functional API.
if all([isinstance(output, base.Node) for output in self.outputs]):
self.hyper_graph = graph.HyperGraph(inputs=self.inputs,
outputs=self.outputs)
# Using input/output API.
elif all([isinstance(output, base.Head) for output in self.outputs]):
self.hyper_graph = meta_model.assemble(inputs=self.inputs,
outputs=self.outputs,
dataset=dataset,
seed=self.seed)
self.outputs = self.hyper_graph.outputs
def test_partial_column_types():
input_node = node.StructuredDataInput(
column_names=common.COLUMN_NAMES_FROM_CSV,
column_types=common.PARTIAL_COLUMN_TYPES_FROM_CSV)
(x, y), (val_x, val_y) = common.dataframe_numpy()
dataset = tf.data.Dataset.zip(
((tf.data.Dataset.from_tensor_slices(x.values.astype(np.unicode)), ),
(tf.data.Dataset.from_tensor_slices(y), )))
hm = meta_model.assemble(input_node, ak.ClassificationHead(), dataset)
for block in hm._blocks:
if isinstance(block, ak.FeatureEngineering):
assert block.input_node.column_types['fare'] == 'categorical'
def _meta_build(self, dataset):
self.outputs = meta_model.assemble(inputs=self.inputs,
outputs=self.outputs,
dataset=dataset)
self._build_network()
def _meta_build(self, dataset):
self.hypermodel = meta_model.assemble(inputs=self.inputs,
outputs=self.outputs,
dataset=dataset,
seed=self.seed)
self.outputs = self.hypermodel.outputs