flsim 0.0.0.5

Federated Learning Simulator (FLSim) is a flexible, standalone core library that simulates FL settings with a minimal, easy-to-use API. FLSim is domain-agnostic and accommodates many use cases such as vision and text.

Model simulation framework simulates model training under FL setting, primarily for ML engineers.

Model simulation framework aims to help ML engineers:

• quickly train an existing model using FL and compare FL training to conventional training
• quickly iterate to pick best model architecture and hyper-parameters for FL

At the same time, we want the model simulation framework to ensure:

• quick training/iteration
• efficient hyper-parameter tuning (Bayesian search)
• developer environment familiar to ML engineers/researchers

Note that this is different from full-stack simulation (a.k.a. Systems Simulation) in "papaya/simulation".

Components

• utils Handy functionalities for simulating Federated Learning such as a class that can simulate event distribution for async FL