C2W1_Assignment - `feature_whitelist`

Course Q&A Machine Learning Engineering for Production(MLOps) Machine Learning Data Lifecycle in Production
1

Hi,

I believe feature_whitelist is replaced with feature_allowlist in the latest version. Might be helpful to update the text to avoid confusion, thanks.

Update: deprecated since v0.28 - Release TensorFlow Data Validation 0.30.0 · tensorflow/data-validation · GitHub

One of the important parameters of the StatsOptions class is feature_whitelist , which defines the features to include while calculating the data statistics. You can check the documentation to learn more about the class arguments.

2

HI @derekchia
welcome to Discourse.
thanks a lot for your valuable feedback.
I have verified that the TFDV version used in the C2W1 assignment is 0.24.1. So the feature_whitelist is still valid.
I have also tried to replace 'feature_whitelist ’ with ‘feature_allowlist’ but it returns an error.

TypeError: init() got an unexpected keyword argument ‘feature_allowlist’

Anyway we will take care of your suggestion for the next versions of the Notebook.
Have you tried to run the notebook in Google Colab? Maybe under Google COlab there is a more updated version of the TFDV framework.
Thanks a lot
BR

3

Hi Fabio,

You are correct. It still works in the C2W1 assignment. I just thought you might want to consider updating it to v1.0.0 onwards since it is a major version update.

I also tried with Google Colab but seems like we need to install it ourselves and there are dependencies issues when doing so. If someone at Google is reading this, please consider adding it as the default package.

Cheers,
Derek

1 Like
4

Hi @derekchia
thanks for your suggestion.
TF dependencies are a little tricky. Furthermore, TF evolves very quickly.
So you were right to point out this possibility.
Thanks a lot
BR

1 Like