sfme / rvae_mixedtypes Goto Github PK

For custom dataset with no categorical columns, the following error will raise when running python core_models/main.py ,

TypeError: reduce() of empty sequence with no initial value from

I think adding a default 0 value in the reduce method should solve the problem.

This resulted in all rows with mismatch in wine_type column. The problem was df_from was dtype str and df_to was dtype float.

Fixed with these two transformation:
' noised_data_df[run_stats["cat_cols_names"]].astype(int).astype(str).apply(lambda x: x.astype('category'))'

in ~ line 476 in `dataset_prep_utils.py'

Probably would be better to fix category type to incoming data to make this more general.