How to fine-tune the RealSR model using a paired dataset？

I only see the use of HR data to fine-tune the model in the Readme of RealSR. I have my own custom dataset of LR-HR pairing. How can I fine-tune based on my paired dataset? I see that most of the content of this document is similar to the fine-tuning document of Real-ESRGAN. Is it the same as Real-ESRGAN?