nvflare.app_opt.pt.scaffold module

class PTScaffoldHelper

Bases: object

Helper to be used with SCAFFOLD components. Implements the functions used for the algorithm proposed in Karimireddy et al. “SCAFFOLD: Stochastic Controlled Averaging for Federated Learning” (https://arxiv.org/abs/1910.06378) using PyTorch. SCAFFOLD-related functions are based on https://github.com/Xtra-Computing/NIID-Bench. See also Li et al. “Federated Learning on Non-IID Data Silos: An Experimental Study” (https://arxiv.org/abs/2102.02079).

get_delta_controls()

get_params()

init(model)

load_global_controls(weights)

model_update(model, curr_lr, c_global_para, c_local_para)

terms_update(model, curr_lr, c_global_para, c_local_para, model_global)

get_lr_values(optimizer: Optimizer)

This function is used to get the learning rates of the optimizer.