DlS Course 2 Week 1 Gradient Check Grader stuck at 80/100

I keep getting 80/100 on the submission for this assignment but I get “All testcases passed!” on every cell. Can someone help me with this? Thanks!

What is the output you get from the gradient_check_n test cell?

The other general point to make is that it’s possible to pass the test cases in the notebook in ways that don’t pass the grader, e.g. if you “hard-code” things to match the test cases. The grader uses different test cases.

1 Like

I get “Your backward propagation works perfectly fine! difference = 1.1890913024229996e-07” since I fixed the backward prop code too. But before that, the grader still outputted 80/100.

Here is my code:

{moderator edit - solution code removed}

Hmmm, that all looks correct to me. Maybe the problem is elsewhere. Please check your DMs …

One weak point of this notebook is that there really isn’t a test for the one dimensional version of gradient_check and there is no “expected value” shown. Here’s what the output should look like with correct code:

Your backward propagation works perfectly fine! difference = 2.919335883291695e-10

What do you see in that section?

Not sure you already fixed this or not… But, here is just for your info.

It looks like the grader checks the output value (difference) of ‘gradient_check()’ function, although a local test does not as Paul mentioned. On the other hand, “check 'gradient_check_n()” seems to check the output value locally. So, one possibility is to look into the function of ‘gradient_check()’, and check the output value.

i got the same issue as you, may i ask how to fix that? i have got correct result from gradient_check_n* test

What was your message for the function ‘gradient_check()’? My value for that function turned out to be wrong.