Unable to remove TypeError: unhashable type: 'numpy.ndarray'

Question

I have just started learning tensorflow coding and I am stuck with very basic code for creating a neural network. Below is the the code:

import tensorflow as tf
import numpy as np    
a = tf.placeholder(shape=(1,1),dtype=tf.float16)
b = tf.placeholder(shape=(1,1),dtype=tf.float16)
y= tf.placeholder(shape=(1,1),dtype=tf.float16)    
addition = tf.add(a, b)    
correct_prediction = tf.equal(addition,y,name=None)    
init = tf.global_variables_initializer()    
with tf.Session() as sess:
    sess.run(init)
    a1=[[0.2]]
    b1=[[0.3]]
    y1=[[0.5]]
    a = sess.run(addition, feed_dict={a: a1, b: b1})
    b = sess.run(correct_prediction, feed_dict={a: a1, b: b1, y: y1})
sess.close()

It is giving me the following error

*Traceback (most recent call last):
  File "D:/NN/pyNN/new_main.py", line 35, in <module>
    b = sess.run(correct_prediction, feed_dict={a: a1, b: b1, y: y1})
TypeError: unhashable type: 'numpy.ndarray'*

I have tried multiple ways for providing the input and output but I end up with the same error. Any help will be highly appreciated

Answer 1

You are assigning the result of session.run() to a. This means a is a numpy array after the first run, overwriting the original definition as a placeholder. Then you are using this array as a key in the dictionary for the second run, which obviously doesn't work.

Renaming the a and b variables within the session context should fix it, e.g.

a_array = sess.run(addition, feed_dict={a: a1, b: b1})
b_array = sess.run(correct_prediction, feed_dict={a: a1, b: b1, y: y1})