This repository was archived by the owner on Oct 1, 2025. It is now read-only.
This repository was archived by the owner on Oct 1, 2025. It is now read-only.
FedSGD with virtual batching #43
Description
🚀 Feature
Motivation
Create a memory efficient client to run FedSGD. If a client has many examples, running FedSGD (taking the gradient of the model based on all of the client's data) can lead to OOM. In this PR, we fix this problem by still calling optimizer.step once at the end of local training to simulate the effect of FedSGD.>