Tensorflow Control Dependencies. Resouce Variables versus normal variables.

tf_control_dep_demo.py

import numpy as np
import tensorflow as tf

def main1():
    x = tf.get_variable("x", initializer=np.float64(0))

    x_id = tf.identity(x)
    x_rd = x.read_value()
    x_p0 = x + 0
    with tf.control_dependencies([x_id, x_rd, x_p0]):
        y = tf.assign_add(x, 1)

    tf_sess = tf.Session()
    tf_sess.run(tf.global_variables_initializer())

    print(tf_sess.run([y, x, x_id, x_rd, x_p0]))


def main2():
    x = tf.get_variable("x", initializer=np.float64(0), use_resource=True)

    x_id = tf.identity(x)
    x_rd = x.read_value()
    x_p0 = x + 0
    with tf.control_dependencies([x_id, x_rd, x_p0]):
        y = tf.assign_add(x, 1)

    tf_sess = tf.Session()
    tf_sess.run(tf.global_variables_initializer())

    print(tf_sess.run([y, x, x_id, x_rd, x_p0]))

if __name__ == '__main__':
    main1()
    tf.reset_default_graph()
    main2()
    # prints:
    #  [1.0, 1.0, 1.0, 1.0, 0.0]
    #  [1.0, 1.0, 0.0, 0.0, 0.0]