Skip to content
#

expert-load-balancing

Here is 1 public repository matching this topic...

This repository contains the core implementation of CRAFT: Cost-Aware Expert Replica Allocation with Fine-Grained Layerwise Estimations. It performs expert-replication benefit analysis and computes a layer-wise expert-replica allocation plan that mitigates expert load imbalance while maintaining high memory efficiency.

  • Updated May 29, 2026
  • Python

Add this topic to your repo

To associate your repository with the expert-load-balancing topic, visit your repo's landing page and select "manage topics."

Learn more