Test a MNIST number

Question

after trained and evaluated a model, I am trying to feed the model with a MNIST test image :

The function Logits fails. X_new_scaled is a 784 long array, scaled from 0.01 to 0.99XXX What does the method logits.eval expect to reiceve? what is the correct way to pass the array?

Answer 1

Your question is a duplicate of TypeError: unhashable type: 'numpy.ndarray' Tensorflow

The point is that you use X twice:

X = mnist["data"]

And

Z = logits.eval(feed_dict={X: X_new_scaled})

So the problem is not X_new_scaled, the problem is that the key of your feed_dict, being X is not a tensor but a a numpy.ndarray.