Tensorflow Developer Exam NLP

[msrraju87]

Hi, I finished your tensorflow course and am ready to take the exam. I stumbled upon a discussion online where a person said he failed the exam as he was unable to save the NLP model as a .h5 extension. Apparently, it is not possible to save the model as a .h5 in case we use TextVectorization and that was precisely what you taught in the course. Can you please let me know of another way we can get this done and still be able to save as a .h5 extension file so that I can attempt the exam.

Thanks,
Raju

[mrdbourke]

Hi @msrraju87,

You don't necessarily need to use the TextVectorization layer if you need to save .h5.

The exam doesn't require you to use any layers that can't be used with .h5.

However, you can also save the TextVectorization layer as part of a model: https://www.tensorflow.org/tutorials/keras/text_classification#export_the_model (though this doesn't allow you to save to .h5, which means it's not necessary to use this layer in the exam).

[abdumuslim]

How can we do that please?
As the embedding layer needs numbers.

[msrraju87]

Hi, I finally passed the exam today. I have to thank you for the wonderful course. I struggled with the time series question and thought I would fail, but luckily I passed it with some guess work. I think the current material in the time series section of your course is not enough to pass this exam as I felt very hard to do the time series question. Can you please add more content on the time series data preparation. It is just a feedback from my side. The other sections were super good however. Thanks for the lovely course and I look forward to taking your data science and pytorch courses soon.

[MiguelCanalGarcia]

@msrraju87 one last question, in the course we learn how to feature extract and fine tune other models from tensorflow hub, is this valid in the exam? Thanks!

[msrraju87]

@juliuschou Thank you very much for the code

[msrraju87]

@MiguelCanalGarcia I didn't use tensorflow hub in the exam. You really don't need to use tensorflow hub to do this exam. I don't know if it is valid or not as I didn't even attempt to use it. I suggest you don't waste your time on that and just concentrate on the non tensorflow hub for the exam.

[MiguelCanalGarcia]

@msrraju87 Thank you for this incredibly valuable information. Having some troubles using tokenizer you mentioned before, I have posted on the discussion of this course one of my codes using it: #430

Sadly I can not reach more than 0.8 in accuracy and I do not know if that is enough. Do you have any github or link where you use it to see what is my error and the proper usage? Thank you very much!

[msrraju87]

@MiguelCanalGarcia I used CNN's and it worked for me. It should work for you as well