ValueError: mismatch of shapes when sampling data for compas dataset about ctgan HOT 2 CLOSED

Hi @bronval thanks for opening this issue, I was able to reproduce this using the same dataset when using CTGAN but not when using CTGANSynthesizer. I suspect that the pre-processing is slightly better with the latter and in general we recommend people use CTGANSynthesizer from the sdv library as it has a significantly better user experience. (Under the hood, CTGANSynthesizer calls out to the original CTGAN library).

Here's a Colab notebook I created where you can see CTGANSynthesizer working perfectly with the same dataset: https://colab.research.google.com/drive/1NJROe2bq28G_IL1Kr6MsP8i5hFuUeM_A?usp=sharing

(Note: You'll need to upload the cox-violent-parsed_filt.csv dataset to Colab in the File view when you run / fork this notebook)

from ctgan.

Hi there @bronval I'm closing this issue for now please feel free to re-open if this issue is still relevant to you :)

from ctgan.