Parallelizing TFF for faster Federated Learning

Hello
I’ve started learning about FL (Federated Learning) and started learning with the provided documentations.

My question is, how can I parallelize the training process of each client so to be asynchronous?
To my understanding, the example mention at this document: Federated Learning for Image Classification performs federated learning on the selected subset of clients but in sequential mode, I’d love to do the same but in parallel.

Any guidance or help, please?

Thanks so much in advance.