GRADED FUNCTION: NER error grader wants log_softmax_v2...how?

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

In the new TensorFlow version (12/2023) of the assignment, the unittest is generating an error for the GRADED FUNCTION: NER as follows:

image
image871×227 17.2 KB

If I change the activation to ‘log_softmax_v2’ the error message is much longer and even less happy than than with ‘log_softmax’.

So, this feels like a Mentor or high-level grade question: How do I provide the ‘log_softmax_v2’ that the grader wants?

Thanks!

2

Hi @jswatdlai

As I understand, you might be passing a string ‘log_softmax_v2’ for the activation= parameter for the keras Dense layer.
I’m really not a big fan of TensorFlow or the current implementation but passing the actual function - tf.nn.log_softmax (note: not a string) as an argument might solve your issue.

Let me know if that helps. Cheers

3

Yep, you’re correct–not using a string but as the function call worked! Thanks!