Seems to be a common issue but none of the threads have helped me. All my tests pass but the grader gives me 80/100.
Below is the test output for UNQ_C5.
I believe I have corrected all I can think of. I have heavily reviewed UNQ_C1-5 and cannot for the life of me figure out what I am doing wrong. I won’t post the code until requested so as to follow the Honor Code.
Since train_step() is pretty simple, most errors are actually hiding in one of the other functions, but aren’t detected in their unit tests.
I found the problem. I had corrected an earlier issue and for some reason the submit tool didn’t register the update. I had used the Restart Kernel and Clear Output option several times as well as Save & Checkpoint but it apparently didn’t take. Note that it was also running and working with the updated version, just not submitting it. I had to completely exit/sign out of Coursera and log back in which reverted it back to the old version.
FYI the earlier error was that I was using tf.square functions rather than ** for scalar values.
paulinpaloalto submitted my code and it worked for him which gave me the idea to just try forcing it to clear.
OK, thanks for the update.