def serving_input_fn():
receiver_1 = tf.compat.v1.placeholder(tf.dtypes.string)
receiver_2 = tf.compat.v1.placeholder(tf.dtypes.string)
receiver_tensors = {
'rec1': receiver_1,
u'rec2': receiver_2,
}
concat = tf.strings.join([receiver_1, receiver_2])
concat2 = tf.identity(concat)
features = {
'feature0': tf.strings.join([concat, concat2], ':'),
u'feature1': tf.constant([1])
}
alternate_tensors = {
'alt_name_1': concat,
'alt_name_2': {
'tensor1': concat,
'tensor2': concat2
}
}
return export_lib.ServingInputReceiver(features, receiver_tensors,
alternate_tensors)
def _serving_input_receiver_fn():
"""A receiver function to be passed to export_saved_model."""
placeholders = {}
time_placeholder = tf.compat.v1.placeholder(
name=feature_keys.TrainEvalFeatures.TIMES,
dtype=tf.dtypes.int64,
shape=[default_batch_size, default_series_length])
placeholders[
feature_keys.TrainEvalFeatures.TIMES] = time_placeholder
# Values are only necessary when filtering. For prediction the default
# value will be ignored.
placeholders[feature_keys.TrainEvalFeatures.VALUES] = (
tf.compat.v1.placeholder_with_default(
name=feature_keys.TrainEvalFeatures.VALUES,
input=tf.zeros(shape=[
default_batch_size if default_batch_size else 0,
default_series_length if default_series_length else 0,
self._model.num_features
],
dtype=self._model.dtype),
shape=(default_batch_size, default_series_length,
self._model.num_features)))
if self._model.exogenous_feature_columns:
with tf.Graph().as_default():
# Default placeholders have only an unknown batch dimension. Make them
# in a separate graph, then splice in the series length to the shapes
# and re-create them in the outer graph.
parsed_features = (
tf.compat.v1.feature_column.make_parse_example_spec(
self._model.exogenous_feature_columns))
placeholder_features = tf.compat.v1.io.parse_example(
serialized=tf.compat.v1.placeholder(
shape=[None], dtype=tf.dtypes.string),
features=parsed_features)
exogenous_feature_shapes = {
key: (value.get_shape(), value.dtype)
for key, value in placeholder_features.items()
}
for feature_key, (batch_only_feature_shape, value_dtype) in (
exogenous_feature_shapes.items()):
batch_only_feature_shape = (
batch_only_feature_shape.with_rank_at_least(
1).as_list())
feature_shape = (
[default_batch_size, default_series_length] +
batch_only_feature_shape[1:])
placeholders[feature_key] = tf.compat.v1.placeholder(
dtype=value_dtype,
name=feature_key,
shape=feature_shape)
batch_size_tensor = tf.compat.v1.shape(time_placeholder)[0]
placeholders.update(
self._model_start_state_placeholders(
batch_size_tensor, static_batch_size=default_batch_size))
return export_lib.ServingInputReceiver(placeholders, placeholders)
def _serving_input_receiver_fn():
"""A receiver function to be passed to export_saved_model."""
times_column = tf.feature_column.numeric_column(
key=feature_keys.TrainEvalFeatures.TIMES,
dtype=tf.dtypes.int64)
values_column = tf.feature_column.numeric_column(
key=feature_keys.TrainEvalFeatures.VALUES,
dtype=values_input_dtype,
shape=(self._model.num_features, ))
parsed_features_no_sequence = (
tf.compat.v1.feature_column.make_parse_example_spec(
list(self._model.exogenous_feature_columns) +
[times_column, values_column]))
parsed_features = {}
for key, feature_spec in parsed_features_no_sequence.items():
if isinstance(feature_spec, tf.io.FixedLenFeature):
if key == feature_keys.TrainEvalFeatures.VALUES:
parsed_features[key] = feature_spec._replace(
shape=((values_proto_length, ) +
feature_spec.shape))
else:
parsed_features[key] = feature_spec._replace(
shape=((filtering_length + prediction_length, ) +
feature_spec.shape))
elif feature_spec.dtype == tf.dtypes.string:
parsed_features[key] = tf.io.FixedLenFeature(
shape=(filtering_length + prediction_length, ),
dtype=tf.dtypes.string)
else: # VarLenFeature
raise ValueError(
"VarLenFeatures not supported, got %s for key %s" %
(feature_spec, key))
tfexamples = tf.compat.v1.placeholder(shape=[default_batch_size],
dtype=tf.dtypes.string,
name="input")
features = tf.compat.v1.io.parse_example(serialized=tfexamples,
features=parsed_features)
features[
feature_keys.TrainEvalFeatures.TIMES] = tf.compat.v1.squeeze(
features[feature_keys.TrainEvalFeatures.TIMES], axis=-1)
features[feature_keys.TrainEvalFeatures.VALUES] = tf.cast(
features[feature_keys.TrainEvalFeatures.VALUES],
dtype=self._model.dtype)[:, :filtering_length]
features.update(
self._model_start_state_placeholders(
batch_size_tensor=tf.compat.v1.shape(
features[feature_keys.TrainEvalFeatures.TIMES])[0],
static_batch_size=default_batch_size))
return export_lib.ServingInputReceiver(features,
{"examples": tfexamples})
def _serving_input_receiver_fn():
receiver = tf.compat.v1.placeholder(
tf.dtypes.float32, shape=[None, 1], name='input')
return export_lib.ServingInputReceiver(
features={'feature': receiver}, receiver_tensors=receiver)
def serving_input_receiver_fn():
return export_lib.ServingInputReceiver(
{'test-features': constant_op.constant([[1], [1]])},
array_ops.placeholder(dtype=dtypes.string))
def serving_input_receiver_fn():
return export_lib.ServingInputReceiver(
{'test-features': tf.constant([[1], [1]])},
tf.compat.v1.placeholder(dtype=tf.dtypes.string))
