Whether server sends updated weights at each round to client or not?.

[sjayarammtech]

What is your question?

Hi, I am using sklearn FL server-clients connection. Using 3 clients. Number of rounds =10 My doubt is,

1. Whether server get weights from all clients at each round, then aggregating the weights using FedAvg, then send the updated weights to all clients at end of that current round (say, round 1). or
2. Server is waiting for all round to finish ,then send the aggregated weight using FedAvg to the clients.

So simply, weather clients get weight update at each round from server or only at final round .

Is the same mechanism used for tensorflow,pytorch models?.

Thanks, S.Jayaram