federated-learing

This frame about federated learning is based on fedprox.

objects

• Simulate more clients in a standalone machine
• Every client has own model

todo

• rewrite get_data.py to make code more beatiful
• implement algorithms such as fedper and so on
• isn't necessary to keep a model for trainer