How to make single prediction

[Wheele9]

Hi

I am new to tensorflow, and I am stuck at your logistic_expression example.
I trained and saved the model, restored it, but I am unable to use it. Can you help me with that?
I tried:

print (sess.run(predict_op, feed_dict={X: teX[66], Y: teY[66]}))
I am expecting a 6 here as output (or a number at least :D) but I got an error:

ValueError: Cannot feed value of shape (784,) for Tensor 'Placeholder:0', which has shape '(?, 784)'

[Waateur]

Hello,
It's feels to me that you made predict_op using a batch input of size : ( None, 784 ) .
But the data in teX are(784,) . You should reshape them in order to be ( 1 , 784 ) .
You can use reshape for a numpy.ndarray or a list comprehension for a float list something like :
[ [x] for x in teX ]

Hope it helped.

[Wheele9]

Hello
I guess I undestood the problem thanks! I solved it like that:

newtx=[ [x] for x in teX ]
newty=[[y] for y in teY]
print (sess.run(prediction, feed_dict={X: newtx[nr], Y: newty[nr]}))

Other question when I load an image from other sources, and reshape it like that. I will not have the label to it, but i have to still give a dummy item to Y right? Is there a more elegant solution? Something like: predicted_label=sess.run(prediction, feed_dict={X:img}) What is the role of Y in th sess.run when the training is done?

p.s

A faster solution is with the numpy, but reshape was not good for me, I had to add a dimension like that and it works now :

img  # shape  : (784,)
img=np.expand_dims(img, axis=0) # shape  : (1, 784)

[nlintz]

You can actually run the predict op without passing in Y:teY since Y isn't part of the predict_op graph. You can just run sess.run(predict_op, feed_dict={X: teX}))

[Wheele9]

Yeah i figured that out, thanks!