How to resume tlt-train faster_rcnn from a specific epoch?

I was training custom data by using Faster_rcnn, the training stopped (System shutdown) in 20/50 epoch, now i want resume training from 20th epoch , how to do it in Faster rcnn.