How to assign tensor values to a numpy array

Question

There are ways to initialize tensor with a numpy array. But is there any way to do the opposite. Meaning initialize a numpy array with a tensor in the graph.

http://stackoverflow.com/questions/36612512/tensorflow-how-to-get-a-tensor-by-name — agent_bean, Jan 30 '17 at 21:16
http://stackoverflow.com/questions/34097281/how-can-i-convert-a-tensor-into-a-numpy-array-in-tensorflow — agent_bean, Jan 30 '17 at 21:21

score 1 · Accepted Answer · answered Jan 30 '17 at 21:52

You have to evaluate the tensor under a session. Assume you have a tensor t defined as

x = tf.Variable(...)
y = tf.Variable(...)

t = tf.add(x, y)

and you want to know its value (given the current x and y). You then just use a session to fetch it:

with tf.Session() as sess:
    numpy_array = sess.run(t)

or, equivalently,

with tf.Session() as sess:
    numpy_array = t.eval(sess)

That should be all there is to it.

How to assign tensor values to a numpy array

1 Answers1