tensorflow initialize variable error

Question

As you all know there are various ways to initialize your variables in tensorflow. I tried some stuff in combination with a graph definition. See the code below.

def Graph1a():
    g1 = tf.Graph()
    with g1.as_default() as g:
        matrix1 = tf.constant([[3., 3.]])
        matrix2 = tf.constant([[2.],[2.]])
        product = tf.matmul( matrix1, matrix2, name = "product")

    sess = tf.Session( graph = g )
    sess.run(tf.global_variables_initializer())
    return product

def Graph1b():
    g1 = tf.Graph()
    with g1.as_default() as g:
        matrix1 = tf.constant([[3., 3.]])
        matrix2 = tf.constant([[2.],[2.]])
        product = tf.matmul( matrix1, matrix2, name = "product")

    sess = tf.Session( graph = g )
    sess.run(tf.initialize_all_variables())
    return product

def Graph1c():
    g1 = tf.Graph()
    with g1.as_default() as g:
        matrix1 = tf.constant([[3., 3.]])
        matrix2 = tf.constant([[2.],[2.]])
        product = tf.matmul( matrix1, matrix2, name = "product")

    with tf.Session( graph = g ) as sess:
        tf.global_variables_initializer().run()
        return product

Why is it so that Graph1a() and Graph1b() won't return product, while Graph1c() does? I thought these statements were equivalent.

Aaron · Accepted Answer · 2017-10-11 16:56:07Z

1

The problem is that the global_variables_initializer needs to be associated with the same graph as the session. In Graph1c this happens because the global_variables_initializer is inside the scope of the with statement of the session. To get Graph1a to work it needs to be rewritten like this

def Graph1a():
    g1 = tf.Graph()
    with g1.as_default() as g:
        matrix1 = tf.constant([[3., 3.]])
        matrix2 = tf.constant([[2.],[2.]])
        product = tf.matmul( matrix1, matrix2, name = "product")
        init_op = tf.global_variables_initializer()

    sess = tf.Session( graph = g )
    sess.run(init_op)
    return product

edited Oct 11, 2017 at 16:56

answered Oct 10, 2017 at 23:04

Aaron

2,3941 gold badge19 silver badges25 bronze badges

Sign up to request clarification or add additional context in comments.

3 Comments

zwep Over a year ago

Too bad, this is not the case. I understand that it looks confusing, but g is already being defined by the with.. as g: I also ran the code with your suggestion of change, and the same situation happened.

Aaron Over a year ago

but the way that you have the identiation g becomes out of scope when the session is initialized

zwep Over a year ago

Okay, as I said before, I have tried your suggestion but it did not change the output. And to check again, I printed the operations of both graph g1 and g in the function.. which are identical. Do you perhaps have any other suggestions? Would you consider this a bug?

Collectives™ on Stack Overflow

tensorflow initialize variable error

1 Answer 1

3 Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

3 Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related