alibaba/FederatedScope — reverse-engineered prompt

Build me an easy to use Python platform for federated learning that lets someone spin up a full training run with a central server and multiple clients, using sensible defaults but still allowing deeper customization later. I want it to work well for research style experiments and practical testing, with clear ways to run standard federated algorithms, compare results, and reproduce benchmark style setups.

Please include support for common machine learning tasks, especially vision, language, and graph cases, plus room for personalized federated learning, vertical federated learning, and federated hyperparameter tuning. It should have simple configs, solid logging and monitoring, example scripts, tests, and beginner friendly docs or notebooks so a new user can get a first experiment running quickly.

Make installation straightforward in a normal Python environment, and ideally also friendly for container based use. A quick start demo is important. If anything is unclear, look up the current project docs online and keep the implementation aligned with them.