Hello! I am running into an interesting issue on this Exercise. My output matches the expected output, but the autograder says otherwise.
I use this equation to calculate dZ1 but I use cache[“Z1”] to reference Z1 inside of
(1-(g’1)^2)
My lab ID is ebqlstqpwmhb
I have carefully checked and confirmed multiple times that the output of my function is the same at the expected output. I have even checked the spaces in the output haha. I have tried restarting the kernel, restarting the server, updating the lab version, and have combed my code for added print statements or code insertions/deletions.
At this point I am wondering if it is not accepting my method of referencing Z1 as I have shown above. It seems to break the normal way in which we are led to reference variables in this course so I wanted to highlight that in my request for help.
Any help would be much appreciated. Please let me know if there are other details that would be helpful to you.
Thanks!
In addition to this, I find that I am unable to submit the assignment while this Exercise if failed. The submit button is not visible. I assume this is an issue since a partial grade of 70% is accepted to pass.
Note that the test input values are just randomly generated, so you can’t assume that A1 = tanh(Z1) for instance. So your code will fail if you use tanh(Z1) in your computation of the derivatives.
I’ve filed this as a bug a while ago, but they have not fixed it yet.
Are you sure you didn’t rename the notebook? If you are working in a renamed copy, then you may not see this button. And if you do, clicking it has a nasty surprise built in: it will not grade your renamed notebook, but the “official” one opened by the “Launch Lab” link.
It can also be a browser or network problem: the “Submit” button is literally the last thing to render when the window gets painted. If something is interfering with your browser or you’re missing some plugins, you may not see the button. Try this:
Kernel -> Restart and Clear Output
Save
Then see if the submit button appears. Note that you can submit to the grader with the output cleared and that’s actually more efficient: the grader just calls your functions, so giving it the output just wastes network bandwidth and cpu for the grader.
Ah, that’s it. Yes I did rename it. Did not realize that it would cause issues. Thanks Paul!
Also, I got to visit Palo Alto back when I interned with Tesla. Really enjoyed that area 
1 Like
Yes, I consider the way “Submit” works with renamed files to be a serious violation of the celebrated UX maxim “The Principle of Least Astonishment”. 
1 Like
Gotcha! I referenced A1 there instead of tanh(Z1) and that worked. I forgot to remember ( ) that A1 = tanh(Z1) and that is far more simple than passing Z1 through the non linear function again. Thanks!
1 Like
Of course your original code is mathematically correct, but (as you say) more expensive than it needs to be. It would be relatively simple for them to rewrite the test case logic so that they don’t randomly generate the A^{[l]} values, but apparently changing the grader is a pretty big hassle. So the bug I filed about this is in the queue, but not fixed yet.
1 Like