In your code, you do not use a discriminator and an additional adversarial loss for better reconstruction.
This is different from what is written in the paper.
Is there another version of code that leverages a well-trained discriminator, or are the checkpoint results based on the official code without discriminator???
In your code, you do not use a discriminator and an additional adversarial loss for better reconstruction. This is different from what is written in the paper. Is there another version of code that leverages a well-trained discriminator, or are the checkpoint results based on the official code without discriminator???