Aditya Sreekar

In the appendix of the paper they mention that they use a prior which is modeled by an LSTM, but in the code here I see that only normal prior...