Tensorflow dynamic/static shapes: Can not convert a int into a Tensor or Operation

In this code, I'm getting the dynamic and static shapes of an input tensor. The problem is that although my Numpy generated array should be considered as a tensor, it does not! Any help will be appreciated!

import tensorflow as tf
import numpy as np


def get_shape(tensor):
"""
Return the static shape of a tensor only when available
"""

static_shape = tensor.shape.as_list()
dynamic_shape = tf.unstack(tf.shape(tensor))

dim = [s[1] if s[0] is None else s[0] for s in zip(static_shape, dynamic_shape)]

return dim


a = tf.placeholder(dtype=tf.float32, shape=[None, 128])

with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
x = np.random.normal(loc=0.5, scale=0.3, size=[150, 128])
shapes = get_shape(a)
print(sess.run(shapes, feed_dict={a: x}))


sess.run(a, feed_dict={a: x})

[<tf.Tensor 'unstack:0' shape=() dtype=int32>, 128]

1 Answer
1

Just change the line

dim = [s[1] if s[0] is None else s[0] for s in zip(static_shape, dynamic_shape)]


to

dim = [s[1] if s[0] is None else tf.constant(s[0]) for s in zip(static_shape, dynamic_shape)]


The thing is that you s[0] in this case refers to int type, because it's a static shape. But here we need a valid tensorflow operation. Using tf.constant(s[0]) instead of s[0] solves the problem.

s[0]

int

tf.constant(s[0])

s[0]

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.