Hello,
I completely stuck with Assignment of Week 1. I made 6 exercises, but when I run the Exercise 6 I get an error:
ValueError: Number of weight elements (1) does not equal the number of sublayers (9) in: Serial_in2_out2[
Select[0,1,0,1]_in2_out4
Can I ask a mentor to check my Notebook?
I am getting the same error, have you solved your problem?
I think I found my issue:
" You can feed these with this syntax to get the model outputs: model((input1, input2))"
I was missing the tuple and directly passing input1 and input2.
@zzwcsong Thank you for the answer. I have changed this point and successfully fixed it. But I started to get the another error (Ex 6):
AssertionError: Tuple must contain an integer and a float number
Does anyone tell me where this problem could be located?
Check if you didn’t forget to convert the prediction to int as in the hint:
# get the next symbol by getting a logsoftmax sample (*hint: cast to an int)
@arvyzukai Thank you! I have fixed the error related to converting to integer and all tests passed.
May I get some assistance with the UNQ_C6 code?
I am failing test_next_symbol UNIT_TEST with the value error:
"ValueError: Number of weight elements (0) does not equal the number of sublayers (9) in: Serial_in2_out2[
Select[0,1,0,1]_in2_out4
"
I found out the source of the error. When I inspected the model.weights calculated few cells above, I received no value. After initializing the weights from the pre-trained NMTAttn model, I was able to pass that check point.
Now I am stuck at this failure:
AssertionError: Expected output: [7283, -9.929085731506348], your output: [140, -0.0002193450927734375]
Any hints? I have tried playing with the output[x,y,:] slice as well as with the logsoftmax_sample temperature. I get always the same error
getting same error as above
Found the problem. Issue was using the pre-trained model above instead of using NMTAttn which was passed in the function.
