Art_Generation_with_Neural_Style_Transfer-UN-GRADED FUNCTION: train_stepQ_C5-

Course Q&A Deep Learning Specialization Convolutional Neural Networks
1

Like many others I have also got this error-
Can someone suggest a final solution for this -
I tried all possible suggestions mentioned in all discussions but nothing seems to be working.
AssertionError Traceback (most recent call last)
in
5 print(J1)
6 assert type(J1) == EagerTensor, f"Wrong type {type(J1)} != {EagerTensor}"
----> 7 assert np.isclose(J1, 10221.168), f"Unexpected cost for epoch 0: {J1} != {10221.168}"
8
9 J2 = train_step(generated_image)

AssertionError: Unexpected cost for epoch 0: 2555.3056640625 != 10221.168

2

Hi prahladpunia,

It’s difficult to say what could be the issue on the basis of what you are reporting here. Could you for starters send me your code for train_step in a direct mail? Then I can have a look.

1 Like
3

I had a DM conversation with Prahlad about this already and the problem is solved. It was just an incorrect implementation of one of the function invocations in train_step.

2 Likes
4

Once again Thanks paul for all the help.

1 Like
5

Thanks - this is resolved.

6

I have a similar problem with the following error:

AssertionError Traceback (most recent call last)
in
5 print(J1)
6 assert type(J1) == EagerTensor, f"Wrong type {type(J1)} != {EagerTensor}"
----> 7 assert np.isclose(J1, 10221.168), f"Unexpected cost for epoch 0: {J1} != {10221.168}"
8
9 J2 = train_step(generated_image)

AssertionError: Unexpected cost for epoch 0: -128130720.0 != 10221.168

I followed the indicated instructions:

  1. In this function you must use the precomputed encoded images a_S and a_C
  2. Compute a_G as the vgg_model_outputs for the current generated image

What could be the problem?

thanks for your help
Juan

7

Notice that your cost value is negative. How could that happen? The costs are positive by definition. You might want to have a look at this thread for an example of one type of mistake that can cause negative cost values.

8

Dear Paulin, thank you very much for the information, with it I solved my problem.
Kind regards,
Juan

9

Hi, Juan.

That’s great news! Thanks for confirming.

Regards,
Paul