how to convert a numpy array in tensor in tensorflow?

Question

I wrote the below code in tensorflow which is a function that I will use to make some calculations based on the input X values

import tensorflow as tf
import math as m

def tf_fn(x):
  pi = tf.constant(m.pi)
  miu=0.0
  o=1.0
  f1=1/(tf.sqrt(2*o**2*pi))
  f2= tf.exp(-((x - miu)**2)/ 2*o**2)
  f3= f1 * f2
  return (f3)
 
x = np.array([0,1,2,3])
tf.print(tf_fn(x))

When I try to print this I get the following error:

InvalidArgumentError                      Traceback (most recent call last)
<ipython-input-70-fba9acd28831> in <module>()
  1 x = np.array([0,1,2,3])
----> 2 tf.print(tf_fn(x))

7 frames
/usr/local/lib/python3.6/dist-packages/six.py in raise_from(value, from_value)

InvalidArgumentError: cannot compute Mul as input #1(zero-based) was expected to be a float 
tensor but is a double tensor [Op:Mul]

My expected output is: [ 0.39894228, 0.24197072, 0.05399097, 0.00443185] I know that the problem is the numpy array that needs to be converted into a tensor. How to do that to get my expected output? Many thanks!!**

Answer 1

Specify in your array creation that you want to use float32 instead of int64 . Numpy defaults to the 64 bits types, but TensorFlow uses float32 for most calculations.

 x = np.array([0,1,2,3], np.float32)

should solve your problem.

---

>>> x = np.array([0,1,2,3], np.float32)
>>> tf_fn(x)
[0.398942292 0.241970733 0.0539909676 0.00443184841]