How can we ensure that the special_key and special_value can be correctly configured for a specific client?

[WLBRich]

What is your question?

IIn the Docs, "Customizing Client Configurations"
In some cases, it may be necessary to send different configurations to individual clients. To achieve this, you can create a custom strategy by extending a built-in one, such as FedAvg:

Example: Client-Specific Configuration

from flwr.server.strategy import FedAvg

class CustomClientConfigStrategy(FedAvg):
def configure_fit(self, server_round, parameters, client_manager):
client_instructions = super().configure_fit(
server_round, parameters, client_manager
)

    # Modify configuration for a specific client
    client_proxy, fit_ins = client_instructions[0]
    fit_ins.config["special_key"] = "special_value"

    return client_instructions

How can we ensure that the special_key and special_value can be correctly configured for a specific client? Relying solely on client_instructions[0] is insufficient to determine which client exactly the [0] represents.

[WilliamLindskog]

Hi @WLBRich,

Thanks for raising this. This has been raised before and we will look at it in April, after our AI Summit which is next week.

We're aiming to reduce the number of issues and will definitely look at this in the coming week(s).

Closing it for now.