synthetic data format is wrong based up on real data -Please help

[Vasanthpravin]

sdv versiom-1.12.0
databricks 13.3 LTS

dp_pandas = df.toPandas()

metadata = SingleTableMetadata()
metadata.detect_from_dataframe(dp_pandas)

synthesizer = GaussianCopulaSynthesizer(metadata=metadata)
synthesizer.fit(data=df_pandas)

synthetic_data = synthesizer.sample(num_rows=50)

display(synthetic_data)
dp_pandas.info(verbose = True, null_counts = False)

Ouptut is coming

Person id PhoneId
sdv-pii-btwry sdv-id-0

both columns are number but its generating sdv like that.Please help

[srinify]

Hi there @Vasanthpravin

When you run metadata.detect_from_dataframe(dp_pandas), SDV does a best-guess effort to automatically infer the metadata (and hence, the sdtypes) for all of your columns.

However, this process isn't perfect and we always recommend double checking the metadata to make sure it matches what you expect. You can display the metadata object to get a read-out of the auto-detected sdtypes:

print(metadata)

Then, you can update the sdtype of multiple columns at once using the update_columns_metadata method from SingleTableMetadata:

metadata.update_columns_metadata(
  column_metadata = {
    'personid': { 'sdtype': 'numerical' },
    'phoneid': { 'sdtype': 'phone_number' }
  }
)

Then you can create your synthesizer object, fit the model, and sample:

synthesizer = GaussianCopulaSynthesizer(metadata=metadata)
synthesizer.fit(data=df_pandas)
synthetic_data = synthesizer.sample(num_rows=50)

[srinify]

Hi there @Vasanthpravin I'm closing out this issue for now, as it seems like there isn't a clear bug here. But let me know if you're still running into the issue or uncover a related bug and we can re-open the issue!