GaussianCopula get_learned_distributions crashes if nothing was learned

[npatki]

Environment Details

• SDV version: 1.17.2.dev0 (but also reproducible on latest SDV version)

Error Description

In some instances, I may have a dataset with columns where nothing gets modeled. For eg, I may just have ID columns, or I may have PII columns that don't get modeled (they are using AnonymizedFaker).

In these instances, fitting and sampling work just fine. But get_learned_distributions causes a crash.

Steps to reproduce

import pandas as pd
from sdv.metadata import Metadata
from sdv.single_table import  GaussianCopulaSynthesizer
import numpy as np

metadata = Metadata().load_from_dict({
    'tables': {
        'table1': {
            'columns': {
                'col_1': {'sdtype': 'id'},
                'col_2': { 'sdtype': 'credit_card_number'}
            },
        }
    }
})
data = pd.DataFrame({
    'col_1': range(100),
    'col_2': range(100)
})

synthesizer = GaussianCopulaSynthesizer(metadata, default_distribution='beta')
synthesizer.fit(data)
synthesizer.get_learned_distributions()

[/usr/local/lib/python3.10/dist-packages/sdv/single_table/copulas.py](https://localhost:8080/#) in get_learned_distributions(self)
    219             )
    220 
--> 221         parameters = self._model.to_dict()
    222         columns = parameters['columns']
    223         univariates = deepcopy(parameters['univariates'])

AttributeError: 'NoneType' object has no attribute 'to_dict'

Expected Behavior

If nothing was learned, then get_learned_distributions should just return an empty dict.

synthesizer.get_learned_distributions()

{}