def test_parse_from_example_list_truncate(self):
with tf.Graph().as_default():
serialized_example_lists = [
EXAMPLE_LIST_PROTO_1.SerializeToString(),
EXAMPLE_LIST_PROTO_2.SerializeToString()
]
# Trunate number of examples from 2 to 1.
features = data_lib.parse_from_example_list(
serialized_example_lists,
list_size=1,
context_feature_spec=CONTEXT_FEATURE_SPEC,
example_feature_spec=EXAMPLE_FEATURE_SPEC)
with tf.compat.v1.Session() as sess:
sess.run(tf.compat.v1.local_variables_initializer())
features = sess.run(features)
# Test dense_shape, indices and values for a SparseTensor.
self.assertAllEqual(features["unigrams"].dense_shape,
[2, 1, 1])
self.assertAllEqual(features["unigrams"].indices,
[[0, 0, 0], [1, 0, 0]])
self.assertAllEqual(features["unigrams"].values,
[b"tensorflow", b"gbdt"])
# For Tensors with dense values, values can be directly checked.
self.assertAllEqual(features["query_length"], [[3], [2]])
self.assertAllEqual(features["utility"], [[[0.]], [[0.]]])
def test_parse_from_example_list_shuffle(self):
with tf.Graph().as_default():
serialized_example_lists = [
EXAMPLE_LIST_PROTO_1.SerializeToString(),
EXAMPLE_LIST_PROTO_2.SerializeToString()
]
# Trunate number of examples from 2 to 1.
features = data_lib.parse_from_example_list(
serialized_example_lists,
list_size=1,
context_feature_spec=CONTEXT_FEATURE_SPEC,
example_feature_spec=EXAMPLE_FEATURE_SPEC,
shuffle_examples=True,
seed=1)
with tf.compat.v1.Session() as sess:
sess.run(tf.compat.v1.local_variables_initializer())
features = sess.run(features)
# With `shuffle_examples` and seed=1, the example `tensorflow` and the
# example `learning to rank` in EXAMPLE_LIST_PROTO_1 switch order. After
# truncation at list_size=1, only `learning to rank` in
# EXAMPLE_LIST_PROTO_1 and `gbdt` in EXAMPLE_LIST_PROTO_2 are left in
# serialized features.
# Test dense_shape, indices and values for a SparseTensor.
self.assertAllEqual(features["unigrams"].dense_shape,
[2, 1, 3])
self.assertAllEqual(
features["unigrams"].indices,
[[0, 0, 0], [0, 0, 1], [0, 0, 2], [1, 0, 0]])
self.assertAllEqual(features["unigrams"].values,
[b"learning", b"to", b"rank", b"gbdt"])
# For Tensors with dense values, values can be directly checked.
self.assertAllEqual(features["query_length"], [[3], [2]])
self.assertAllEqual(features["utility"], [[[1.]], [[0.]]])
def _predict(serialized):
features = data.parse_from_example_list(
serialized,
context_feature_spec=self._context_feature_spec,
example_feature_spec=self._example_feature_spec,
mask_feature_name="mask")
scores = dnn_model(inputs=features, training=False)
return {"predictions": scores}
def _predict(serialized):
features = data.parse_from_example_list(
serialized,
context_feature_spec=context_feature_spec,
example_feature_spec=example_feature_spec,
size_feature_name="example_list_size")
scores = ranker(inputs=features, training=False)
return {"predictions": scores}
def predict(serialized_elwcs: tf.Tensor) -> Dict[str, tf.Tensor]:
"""Defines predict signature."""
features = data.parse_from_example_list(
serialized_elwcs,
context_feature_spec=self._context_feature_spec,
example_feature_spec=self._example_feature_spec,
mask_feature_name=self._mask_feature_name)
outputs = self._model(inputs=features, training=False)
return _normalize_outputs(tf.saved_model.PREDICT_OUTPUTS, outputs)
def test_parse_example_list_with_sizes(self):
with tf.Graph().as_default():
serialized_example_lists = [
EXAMPLE_LIST_PROTO_1.SerializeToString(),
EXAMPLE_LIST_PROTO_2.SerializeToString()
]
# Padding since list_size 3 is larger than 2.
features = data_lib.parse_from_example_list(
serialized_example_lists,
list_size=3,
context_feature_spec=CONTEXT_FEATURE_SPEC,
example_feature_spec=EXAMPLE_FEATURE_SPEC,
size_feature_name=_SIZE)
with tf.compat.v1.Session() as sess:
sess.run(tf.compat.v1.local_variables_initializer())
features = sess.run(features)
self.assertAllEqual(features[_SIZE], [2, 1])
def test_parse_from_example_list_static_shape(self):
with tf.Graph().as_default():
serialized_example_lists = [
EXAMPLE_LIST_PROTO_1.SerializeToString(),
EXAMPLE_LIST_PROTO_2.SerializeToString()
]
feature_map_list = []
for list_size in [None, 100, 1]:
feature_map_list.append(
data_lib.parse_from_example_list(
serialized_example_lists,
list_size=list_size,
context_feature_spec=CONTEXT_FEATURE_SPEC,
example_feature_spec=EXAMPLE_FEATURE_SPEC))
for features in feature_map_list:
self.assertAllEqual(
[2, 1], features["query_length"].get_shape().as_list())
for features, static_shape in zip(feature_map_list, [
[2, 2, 1],
[2, 100, 1],
[2, 1, 1],
]):
self.assertAllEqual(static_shape,
features["utility"].get_shape().as_list())
