Newshot

Fuse-All MoE Inference Kernel with Load Balancing on TPUs

Summary

arXiv:2611.17857v1 Announce Type: new Abstract: Mixture-of-Experts (MoE) architectures activate only a subset of experts for each token, allowing large language models (LLMs) to scale parameter capacity without a proportional increase in computation per token. However, intermediate data movement and uneven expert loads limit the efficiency of MoE inference on TPUs. To address these bottlenecks, we introduce Fuse-All MoE, a TPU kernel design that jointly optimizes expert computation, output communication, and load balancing. The kernel fuses token collection, expert projections, gated activation, and output writes, retaining expert computation intermediates on chip. Fixed token-slot destinations allow scheduling and expert placement to change while preserving routing semantics. In prefill-decode (PD) disaggregated serving, Fuse-All MoE balances EP16 prefill workloads through scheduling and expert redistribution, while symmetric EP8TP2 shares active expert computation between the two Ten…

This source only provides a headline — there is no body to show. Use “Read the original” below to read it at the source.

Published by arXiv cs.LG
This site only summarises and aggregates; copyright stays with the original author. Headlines and summaries follow the source — where they differ, the original wins.
Read the original ↗