What is the function, def load_synthetic_data(args), used for?

I found the part of FedML paper that mentions synthetic data, and then I found this function in data_load.py. But due to my lack of skills in coding I didn't understand much about this function, load_synthetic_data. Does it generate the NON-IID data on each client from the existing dataset using the synthetic data method, or does it simply perform the splitting operation.

If the synthetic data method is utilized, does FedML provide the ability to generate synthetic data based on the dataset I provided？