b_e = tf.get_variable("b_e", [batch_size, lstm_width], initializer=init) # B x L (bias matrix) I guess the shape of b_e should be [lstm_width, 1] or [1, lstm_width]
b_e = tf.get_variable("b_e", [batch_size, lstm_width], initializer=init) # B x L (bias matrix)
I guess the shape of b_e should be [lstm_width, 1] or [1, lstm_width]