About the src/layers/rnn_cell.py

[JLiu-1]

In the call() function of EideticLSTMCell class, on line 210 and line 215 there are:
output_gate = tf.tanh(o_x + o_h + o_c + o_m)
output = tf.tanh(memory) * tf.sigmoid(output_gate)
However, according to the paper, I think these 2 lines should be:
output_gate = tf.sigmoid(o_x + o_h + o_c + o_m)
output = tf.tanh(memory) * output_gate
Are there any explanations that there's and additional tanh activation on the (o_x + o_h + o_c + o_m)?