jeff-hou / urop-adversarial-feature-matching-for-text-generation Goto Github PK

in discriminator.py
def discriminate() has self.cnn_out which has not been initialized but it is working i dont know how?
and what exactly is discriminate() doing?
discriminate() is able to compute cnn_out but at last it is returning self.out() which calls the out()
cnn_out is again passed to the fully connected layer but is cnn_out any way related to discriminate() which computed it and how exactly does "self.cnn_out = tf.concat([self.cnn_out, c], axis=1)" in discriminate() work as self.cnn_out has not been declared or intialized?

please help me to understand in detail.

Hello Jeff,

Thanks for sharing your material. I am currently reading the same paper and was searching for the source code when I stumbled across your repo. May I ask where you got the source code from? Is this from the original authors? I would like to verify this before looking at their code in more depth.

Thanks for your help and best!
Pete