Fix germancredit.json, add two additional executable scripts, set PYTHONPATH to CTGAN

tandriamil commented 3 years ago

Fix the type of the attributes in the metadata file germancredit.json.

Also ignore every folder that begins with venv instead of only venv3.7 as lazy people (like me) just name their virtual environment venv.

tandriamil commented 3 years ago

Additionally, I added two scripts in a new executables folder:

generate_metadata_file.py allows to generate a metadata json file from a dataset in csv format (the example for now comes from the adult dataset). This would meet the demand of a person asking for such script in Issue #6.
generate_synthetic_dataset.py allows to simply generate a synthetic dataset (adapted from the *_cli.py scripts).
I use loguru to display logs and added it to the requirements.

I added in the README how to set the PYTHONPATH to CTGAN to avoid import errors (see Issue #3).

Feel free to only take parts of the proposed edits directly from my fork, thank you.

tandriamil commented 3 years ago

Add a fix to the call of the CTGANSynthesizer constructor that contained a metadata parameter whereas it does not accept such parameter.

laurent-girod commented 3 years ago

Nice! Thank you for your contribution!

spring-epfl / synthetic_data_release