Created by: CrossLee1

In the demo of sentiment, I see the stack lstm is used for this classification task (this file).

I have some questions about the learning rate. In line 105, the para_attr is a list, which contains the learning rate of fc_layer and lstm_layer. In line 123, the fc_layer uses the param_attr to get the learning rate, but the following lstm does not have a param_attr. Then could the second element of the para_attr be conveyed to lstm? How?