C5_W4_A1_Transformer_Subclass_v1 grader output

All tests passed but grader output shows as 50/100. Is there some issue with this exercise?

Hi @souravmodi22

I am not aware of any issue with this assignment lately. Passing all the unit tests provides some confidence of your code, but these are not the full test. The auto grader must have picked up some problem with your code.

1 Like

What messages do you get from the grader? They’re available just beneath your score.

I got the below messages from the grader based on which I observed that Assertion error is in the unit tests. Currently Jupyter notebook shows no error but the grader output is same.
Code Cell UNQ_C1: Function ‘get_angles’ is correct.
Code Cell UNQ_C2: Function ‘positional_encoding’ is correct.
cell_UNQ_C3 function scaled_dot_product_attention failed test 2(with mask)
Code Cell UNQ_C4: Function ‘EncoderLayer’ is correct.
Code Cell UNQ_C5: Function ‘Encoder’ is correct.
cell_UNQ_C6 function DecoderLayer failed test 1
cell_UNQ_C7 function Decoder.call failed test 1
cell_UNQ_C8 function Transformer.call failed test 1

Hi @souravmodi22 ,

It is reported that the scaled_dot_product_attention() failed the test when the mask is on. So check if the mask is added to the scaled tensor.
Once this problem is sorted, we can then move on to solve the rest.