naiyarachuhdry/federated-learning-flower-pytorch ? reverse-engineered prompt

Build me a simple federated learning demo using PyTorch and Flower.

I want a small project that trains a model on CIFAR 10, splits the data across multiple clients, and shows how the clients and server work together to improve the model over several rounds. Please make it easy to run locally in simulation mode first, with a clear way to change things like number of rounds and learning rate from a config or command line. If it makes sense, also keep it ready to run in a real deployment setup later without having to rewrite the code.

Please include the basic client logic, server logic, and model and data loading code, plus whatever setup is needed so I can install dependencies and start it quickly. Keep it beginner friendly, with sensible defaults and a README that explains how to launch the app. If you need to check current Flower docs online for the latest recommended setup, go ahead.