def _example_serving_input_fn():
feature_spec = tf_lib.get_feature_spec(model_name, ctx, training=False)
example_bytestring = tf.placeholder(shape=[None], dtype=tf.string)
feature_scalars = tf.parse_single_example(example_bytestring, feature_spec)
features = {key: tf.expand_dims(tensor, -1) for key, tensor in feature_scalars.items()}
return tf.estimator.export.ServingInputReceiver(
features=features, receiver_tensors={"example_proto": example_bytestring}
)
def generate_example_parsing_fn(model_name, ctx, training=True):
model = ctx.models[model_name]
feature_spec = tf_lib.get_feature_spec(model_name, ctx, training)
def _parse_example(example_proto):
features = tf.parse_single_example(serialized=example_proto, features=feature_spec)
target = features.pop(model["target_column"], None)
return features, target
return _parse_example
