Le-Xiaohuai-speech / DPCRN_DNS3

Implementation of paper "DPCRN: Dual-Path Convolution Recurrent Network for Single Channel Speech Enhancement"
188 stars 41 forks source link

why different size can be assign ? #4

Closed l2009312042 closed 3 years ago

l2009312042 commented 3 years ago

it is a good job , in the file stateful_modules.py,there is a assign code as follow: self.update_h = h_state.assign(tf.expand_dims(inter_LSTM_h,axis = 0),read_value = False)

the h_state shape is (b,f,c) ,the parentheses part's shape is (1,b*f,c),why can be assign to h_state, the shape is not equal?

iknow why in this place b =1 so the shape is the same