Architecture W1 A3 multi-layered

Hi,

I’m wondering about the architecture of the Jazz improvisation project. Why put the same LSTM layer in a for loop, in a way that resembles that of building multi-layer LSTM, when LSTM is already a recurrent layer and no for loop is needed? Is it only for reading the input in parallel?

Thank you

Which function and for-loop are you referring to specifically?

I am referring to the for loop inside the djmodel function. To execute a step in the LSTM cell we use a for loop, save the values we pass to the next step back to the LSTM cell. In this way we create many LSTM layers formed by the same LSTM, when everything would be done automatically without the for loop. Am I mistaken?