It turns out that generate_samples_main.py was breaking because the GAN in gan_training_main.py was ignoring the autoencoder param and so was not in fact training an AE-GAN. This happened because the user param path_autoencoder had been renamed to autoencoder but this was not changed in the init_gan function nor in the generate_samples_main script.
This is in response to issue #86.
It turns out that
generate_samples_main.py
was breaking because the GAN ingan_training_main.py
was ignoring the autoencoder param and so was not in fact training an AE-GAN. This happened because the user parampath_autoencoder
had been renamed toautoencoder
but this was not changed in theinit_gan
function nor in thegenerate_samples_main
script.This PR made these changes to fix the issue.