W2_A2_Failing Test on a hidden cell

Course Q&A Deep Learning Specialization Neural Networks and Deep Learning
1

Hi,
How can I see the corrected version of my assignment once I received my score in order to understand my mistakes ?
The “Grader Output” doesn’t seem very explicit e.g “Tests failed on 1 cell(s)! These tests could be hidden. Please check your submission.” is mentioned for my “Practice Programming Assignment” of Week 2.
Thanks in advance
Regards,

2

Hi @Didier_Heck ,

Before submitting your assignment, could you do the following from the top of the menu bar:

  1. from the kernel tab, click restart & clear all output
  2. from the cell tab, click run all
  3. checking there is no error before hitting the submit button
1 Like
3

Hi Kic,
Just did what you said.
Only cell 6 is in error because it’s an example why we don’t use math but numpy instead.

image
image873×361 18.6 KB

But I only had 88% to my assignment.

Regards,

Didier

4

Hi @Didier_Heck ,

math.exp() only works on scalar values. But np.exp() can operate on array of values. So when passing an array to math.exp(), it would raise an error as reported here.

5

Unfortunately the grader sometimes can’t tell us which function is the one that failed the tests. We have asked the course staff about this a number of times and it’s apparently not easy to fix. One experiment to try would be to comment out the call to basic_sigmoid with the array input which is failing and then “Save” and “Submit” again. My guess is that will still not help, because there is probably a real error someplace else.

If we can’t get any better information from the grader in a case like this we may have to examine your code ourselves. There are no “official solutions” or any way for the grader to show you the correct answers. I will send you a DM about how to proceed with that.

6

Hi @Didier_Heck ,

When you click the cell tab → run all, the kernel would stop at cell 6 because of using math.exp() with x, a vector, as input argument. You need to run the rest of the code cells to the end of the assignment. The grader would not take into account of cell 6 as this is designed for you to see the difference between math.exp() and np.exp().
I have just rerun my code and submitted it without any issue and got 100% of the marks.

7

Thanks Kic,

But this doesn’t tell me why I received 88% as this cell was not an exercise but just an explanation identical to the one you gave here under!

8

To close the loop on the public thread, we had a DM conversation to examine the source code and the problem turned out to be a “hard-coding” error in the softmax function that assumed that the input matrix would always have 2 rows. It worked in the notebook test, but failed the grader. The “take away” point here is that it’s always a mistake to hard-code any assumptions about the inputs unless you literally have no choice and they specifically tell you to do that.

Of course it’s still a question why the grader couldn’t be bothered to tell us which function was wrong, but (as I mentioned earlier) that will just have to remain a mystery. The only good news is that there is a way to proceed when you get hit with that conundrum.