Fix Fenced Doc-String #782

[atharvapurdue]

Following changes were made in Doc-String Examples -

• Add models to the list in docstring_test.py
• Import from the root of the library
• Fix typo in f_net_classifier.py
• add [MASK] in Vocaburay
• remove num_heads parameter

The number of failed examples has gone down from 5 to 1, however I am getting 1 error in File "/content/keras-nlp/keras-nlp/keras-nlp/keras_nlp/models/f_net/f_net_masked_lm.py", line 73, in keras_nlp.models.f_net.f_net_masked_lm.FNetMaskedLM-

Node: 'f_net_masked_lm_1/f_net_backbone_3/segment_embedding/embedding_lookup' indices[0,3] = 4 is not in [0, 4) [[{{node f_net_masked_lm_1/f_net_backbone_3/segment_embedding/embedding_lookup}}]] [Op:__inference_train_function_458156]
Might need help from @abheesht17 , or any other developer who can assist.
Thanks!
Fixes keras-team/keras-nlp #782

[abheesht17]

Thanks, @atharvapurdue! The reason why it's throwing an error is because the default value of num_segments is 4 for FNet. Hence, this throws an embedding lookup error: https://github.com/keras-team/keras-nlp/blob/master/keras_nlp/models/f_net/f_net_backbone.py#L126-L131. The best way to resolve it is to just change

        "segment_ids": tf.constant(
            [[1, 0, 0, 4, 0, 6, 7, 8]] * 2, shape=(2, 8)
        ),

to values less than 4. Just change it to something like [[0, 0, 0, 1, 1, 1, 0, 0]] * 2.

[atharvapurdue]

Yes! @abheesht17. I thought so, but wasn't sure. I will make the necessary changes. Thank you!

[atharvapurdue]

Done ! @abheesht17 , Thanks for your help!
You can see the doc_string_test passing all tests here - https://colab.research.google.com/drive/1xVlUGJIDSH44bF851HZ01yD24ocJmul5#scrollTo=8DErjytmUTv_

[mattdangerw]

Looks good, a few minor comments.

[mattdangerw]

I think these three additions should be in the "base classes" block a few lines up.

[atharvapurdue]

Made the required changes @mattdangerw. Please have a look

[mattdangerw]

Thank you! Works great!