CTGAN icon indicating copy to clipboard operation
CTGAN copied to clipboard

add progress bar

Open AndresAlgaba opened this issue 3 years ago • 3 comments

This PR is a proposal for issue #179. A previous PR #181 was created but closed; I am unsure why. It requires adding the tqdm module to the imports; not sure whether this is something you want to support?

AndresAlgaba avatar Jul 25 '22 11:07 AndresAlgaba

Hi @AndresAlgaba, appreciate your time to look into this and write a PR. I believe there was some additional discussion about whether to add support for a progress bar (and tqdm) in this library or the overall SDV library.

The CTGAN library is meant for the core ML algorithm while the SDV library contains user-friendly wrappers. It may be more advantageous to think through and write a broader solution in the SDV library that works for all tabular models -- not just CTGAN.

I'll still have to discuss with some of the other SDV maintainers so we can leave this PR open for now.

npatki avatar Jul 25 '22 20:07 npatki

Hi @npatki, thanks for the feedback! It may indeed be something that belongs in the SDV library. I found: sdv-dev/SDV#913, for example. One obstacle I see is that the fit method is called in the SDV wrapper, whereas the sample methods are overwritten, and that, therefore, it is only possible to add it here.

AndresAlgaba avatar Jul 27 '22 13:07 AndresAlgaba

Hi @AndresAlgaba, makes sense to me. We may also have to review the API a bit, as some users may not want to see this verbosity as a default (without the ability to turn it off). Currently, the verbose parameter is set to print out the discriminator vs. generator loss for each epoch.

Note that we are currently the process of deprecating our old, CLA assistant bot and moving over to a new Contributor Licensing Agreement.

To request our new Contributor Licensing Agreement, we ask that you please fill out the information in the form: https://bit.ly/sdv-cla-form. Once we receive your submission, we'll send you a PDF to the email that you give us.

You will only have to do this once for contributing to any of the libraries in the SDV ecosystem. Thanks for bearing with us!

npatki avatar Mar 13 '23 21:03 npatki