search

sdv-dev / CTGAN

Conditional GAN for generating synthetic tabular data.
Other
1.23k stars 280 forks source link

Investigate convergence warnings & silence if they can be safely ignored #202

Open sparshgarg23 opened 2 years ago

sparshgarg23 commented 2 years ago

Environment Details

Please indicate the following details about the environment in which you found the bug:

Error Description

While running CTGAN demo on my system,I encountered the following warning ConvergenceWarning: Initialization 1 did not converge. Try different init parameters, or increase max_iter, tol or check for degenerate data. warnings.warn('Initialization %d did not converge.

Steps to reproduce

Use pip install ctgan Execute demo code on windows operaitng system

python ctgan_demo.py
npatki commented 2 years ago

Hi @sparshgarg23, the convergence warnings may indicate that the model having some trouble with mathematical optimization for that particular epoch. I see this frequently. It is not indicative of a bug and it usually does not correspond to any degradation in synthetic data quality.

Let's repurpose this issue as a feature request to investigate such warnings and figure out if we can safely silence them so they are not alarming during the fitting phase.