Q1. it seems that you call tf.get_variable_scope().reuse_variables() function many times in the scope "teller" in rollout.py, in my understanding the first time you call this function makes the reuse flag True in the whole scope, why call it many times?
Q2. are the weights of the word embedding in discriminator different from that in the generator? can i reuse the word embedding of generator in the discriminator?
Q1. it seems that you call tf.get_variable_scope().reuse_variables() function many times in the scope "teller" in rollout.py, in my understanding the first time you call this function makes the reuse flag True in the whole scope, why call it many times? Q2. are the weights of the word embedding in discriminator different from that in the generator? can i reuse the word embedding of generator in the discriminator?