The self.reshape_features is empty when working on petfinder dataset

[zhisbug]

Describe the bug
This issue is related with #1134.

I tried to replace the reshape_features with features to workaround the bug caused by IteratableBatcher (see #1134 for the detailed descriptions about the issue), but when working on the petfinder dataset, the self.reshapes_features is empty after applying the following code between L35-L39 of parquet.py:

self.reshape_features = {
    feature[PROC_COLUMN]: list((-1, *training_set_metadata[feature[NAME]]['reshape']))
    for feature in features
    if 'reshape' in training_set_metadata[feature[NAME]]
}

To Reproduce
Use my experiment script on petfinder dataset here: https://github.com/zhisbug/ludwig-petfinder/blob/master/petfinder-adoption-prediction/train.py

Try to print self.reshaped_features and you will see it is an empty dict.

Expected behavior
A clear and concise description of what you expected to happen.

Screenshots
If applicable, add screenshots to help explain your problem.

Environment (please complete the following information):

• OS: [e.g. iOS]
• Version [e.g. 22]
• Python version
• Ludwig version

Additional context
Add any other context about the problem here.

[zhisbug]

See PR#1136 my workaround on this

[tgaddair]

Hey @zhisbug, are you running with image features? If not, then it's expected that rehshape_features should be empty (as there are no features that need to be reshaped when loaded from the Parquet file). Is this still an issue after applying #1138?

[zhisbug]

Yes!