TextVectorizer and .h5 format

[sodeso-nl]

In the Tensorflow Developer Exam we will need to save the models in .h5 format, how do we do that for NLP models were we are using the TextVectorizer? Since this layer does not support saving? Will this become clear when doing the exam or is there some trick that is not mentioned in the course?

Or do we need to use a different construction? I mean, we can use the TextVectorizer ouside the model to pre-process all the training data and validation data. But how would that work when submitting the model? Google is probably testing the model with their own sentences, how do they know how to vectorize them to the same numbers.

[LubuntuSharp]

facing the same issue, where you able to find a solution ?
thanks,

[sodeso-nl]

So i did the Exam and altough the TensorFlow xam is not available anymore i cannot go into details due to the NDA, however i can point out that in normal circumstances you can for example have the TextVectorizer outside of the model and vectorize the text for example in your data pipeline in a .map function. Then it is not part of the model and it will not cause any issues.

[LubuntuSharp]

Thanks a lot, you help is greatly appreciated!
Regards,

[Mohsenzergani]

use this command
checkpoint_path = "ten_percent_model_checkpoints_weights.h5/checkpoint.weights.h5"