How do you create a bidirectional convolutional LSTM network in Pytorch?

This Github shows how to make convolutional LSTM in Pytorch but what do I have to do to make it bidirectional? From what I can understand to be bidirectional you have to calculate another forward pass but starting in the opposite direction and concatenate the final result of those 2 directions together? How would you implement this in pytorch?

I found the above article haven’t read it yet, hopefully I can translate RNN to LSTM but it doesn’t hurt to ask here directly.