Assignment (pass by reference) in TensorFlow

gist.py

# This note is to show that TensorFlow objects are shallow copied:
# a deep copy is not made when a class contains another class.
#
# This is relevant for variational parameters. For example, Inference
# holds Variational, we train variational parameters via
# inference.run(). We can use the original variational object as it will
# have the trained parameters, and not just resort to using
# inference.variational.
# (This is probably a feature in Python in general.)
from __future__ import print_function
import tensorflow as tf

class ClassWithVar:
    def __init__(self):
        self.var = tf.Variable(tf.random_normal([1]))
        self.non_var = tf.constant(0.0)

class Container:
    def __init__(self, x):
        self.x = x

sess = tf.InteractiveSession()

x = ClassWithVar()
y = Container(x)

init = tf.initialize_all_variables()
sess.run(init)

print(x.var.eval())
print(y.x.var.eval())
print(x.non_var.eval())
print(y.x.non_var.eval())

y.x.var += tf.constant(1.0)
y.x.non_var += tf.constant(1.0)

print(x.var.eval())
print(y.x.var.eval())
print(x.non_var.eval())
print(y.x.non_var.eval())