def setUp(self):
super(DNNRankingNetworkTest, self).setUp()
self.context_feature_columns = _context_feature_columns()
self.example_feature_columns = _example_feature_columns()
self.features = _features()
self.network = dnn.DNNRankingNetwork(
context_feature_columns=self.context_feature_columns,
example_feature_columns=self.example_feature_columns,
hidden_layer_dims=["10", "10", "10"],
activation=tf.nn.relu,
dropout=0.5)
def test_call_empty_example_hidden_layer_dims(self):
with self.assertRaisesRegexp(
ValueError, r"example_feature_columns or "
"hidden_layer_dims must not be empty."):
dnn.DNNRankingNetwork(
context_feature_columns=self.context_feature_columns,
example_feature_columns=self.example_feature_columns,
hidden_layer_dims=[],
activation=tf.nn.relu,
dropout=0.5,
name="dnn_ranking_network")
def test_call_none_context_feature_columns(self):
network = dnn.DNNRankingNetwork(
context_feature_columns=None,
example_feature_columns=self.example_feature_columns,
hidden_layer_dims=["10", "10"],
activation=tf.nn.relu,
dropout=0.5,
name="dnn_ranking_network")
logits = network(inputs=self.features,
mask=[[True, False], [True, True]])
self.assertAllEqual([2, 2], logits.get_shape().as_list())