UNQ_C8 - fails grader

Course Q&A Natural Language Processing NLP with Attention Models
1

Hello,

My rouge score is getting a gade of zero. I get “All tests passed” with an output of 0.857, but when I run it through the grader, it complains about the output being 0.857. Please see below:

Failed simple_test_check.
Expected:
0.7868852459016393,
but got:
0.8571428571428571.

I have checked this in depth with print statements and I got a Precision score of .75 (3 out of 4 matched) and recall score of 1.0 (3 out of 3 matched). When I calculate the 2PR/(P+R), it is indeed .857.

My lab ID is lblxejyn

Thanks!

2

Hi @David_Simmonds

The most common case in these type of errors is when global variable is used - your notebook might get the correct result using the variable the grader doesn’t know about. Please double check all your variables and the way you use them.

If you still have the problem, please private message me your notebook and I will look into it.

Cheers

3

Hello Arvyzukai,

I don’t think I am doing that. I only use the parameters of the function. I would deeply appreciate if you would glance at my notebook. My lab id is lblxejyn

Thanls,
David

4

I cannot access your notebook by lab_id. As far as I understand, mentors cannot access learners’ assignments by lab_id, only staff members can do that.

5

I’m having the same issue, all tests passed but the grader failed. Could you help me? I dont’ have any clue wich variables i’m messing up with.

6

You have to make sure that you are not using global variables within your functions. Also, make sure to remove all print statements. This worked for me.

7

I have the same problem but blaming "global variable " can’t be right. This function doesn’t use any global variables. Also there were no print in this function. Please help. My ID is PZ2004

8

On the contrary @PZ2004 , the common problems:

  • do not use global variable weighted_avg_overlap but instead use local score_fn;
  • do not use global variable model but instead use local NMTAttn;

Other are less common but maybe one of these solves your problem?