Hiii Osman,

Thank you for your clear implementation, its really helpful!
My question is there anyway that we can get the generator's output as Text?
Is it possible to see what exactly it is the generating?

Although the loss gets smaller when training, in testing model only predict one class for the entire test set.

The only exception was when I was training with my data (balanced binary classification data) after ossilating between the two classes through epochs (always achieving 50% on dev), in epoch 20 the model finally predicted different classes for different samples and achieved a score around 60% on development set. I could not replicate this phenomenon again.

Hi, thanks for the excellent implementation of pytorch version GAN-BERT! I have 3 questions for the code:

1. Generator in GAN-BERT is a MLP, in model.py it makes the following transformation of noise's dimension: 100 -> 512 -> 512, why just only 100 -> 512?
2. Why should we import parameters of TensorFlow version in def convert_to_tf_param_name and def get_weights_from_tf of model.py?
3. In general GAN, we often train discriminator for several times then train generator once in one epoch. But in ganbert.py, it seems you train discriminator only once then train generator once. Will this operation cause the problem mentioned in issue#1?

Thanks!