search

tsrobinson / SyGNet

Synthetic data using Generative Adversarial Networks
GNU General Public License v3.0
11 stars 2 forks source link

Add critic score to output during training #11

Closed tsrobinson closed 2 years ago

tsrobinson commented 2 years ago

With our switch to Wasserstein loss, we know the the critic value is an informative measure of model quality (thanks @antndlcrx for pointing this out!) Therefore we should report this during training as part of the per-epoch tcdm.

I also think we could have this as a default function for the tune() function, re. #7.

tsrobinson commented 2 years ago

I'll take this one and get it implemented today!