Pith. sign in

REVIEW 6 cited by

HeterMoE: Efficient Training of Mixture-of-Experts Models on Heterogeneous GPUs

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2504.03871 v1 pith:DWDDCERY submitted 2025-04-04 cs.DC cs.LG

HeterMoE: Efficient Training of Mixture-of-Experts Models on Heterogeneous GPUs

classification cs.DC cs.LG
keywords gpushetermoeexpertoldertrainingattentionmodelscompared
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
read the original abstract

The Mixture-of-Experts (MoE) architecture has become increasingly popular as a method to scale up large language models (LLMs). To save costs, heterogeneity-aware training solutions have been proposed to utilize GPU clusters made up of both newer and older-generation GPUs. However, existing solutions are agnostic to the performance characteristics of different MoE model components (i.e., attention and expert) and do not fully utilize each GPU's compute capability. In this paper, we introduce HeterMoE, a system to efficiently train MoE models on heterogeneous GPUs. Our key insight is that newer GPUs significantly outperform older generations on attention due to architectural advancements, while older GPUs are still relatively efficient for experts. HeterMoE disaggregates attention and expert computation, where older GPUs are only assigned with expert modules. Through the proposed zebra parallelism, HeterMoE overlaps the computation on different GPUs, in addition to employing an asymmetric expert assignment strategy for fine-grained load balancing to minimize GPU idle time. Our evaluation shows that HeterMoE achieves up to 2.3x speed-up compared to existing MoE training systems, and 1.4x compared to an optimally balanced heterogeneity-aware solution. HeterMoE efficiently utilizes older GPUs by maintaining 95% training throughput on average, even with half of the GPUs in a homogeneous A40 cluster replaced with V100.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 6 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. HetRL: Efficient Reinforcement Learning for LLMs in Heterogeneous Environments

    cs.DC 2025-12 unverdicted novelty 7.0

    HetRL delivers up to 9.17x higher throughput for LLM RL training on heterogeneous GPUs by using hybrid and ILP-based schedulers to solve a joint optimization problem over computation and data dependencies.

  2. Simulating Unified Tensor Resharding in heterogeneous AI systems

    cs.DC 2026-06 unverdicted novelty 6.0

    Xsim is a heterogeneity-aware simulator for distributed LLM training supporting load balancing, customized collectives, tensor resharding, and pluggable network simulation, reporting under 5% error in training time pr...

  3. Diagnosing Overhead in Dispatch Operations: Cross-architecture Observatory

    cs.DC 2026-05 unverdicted novelty 6.0

    DODOCO measurements show MoE routing imbalance is intrinsic to architecture and real text, not correctable by EP scaling or represented by mock tokens, forming two persistent Gini bands.

  4. Diagnosing Overhead in Dispatch Operations: Cross-architecture Observatory

    cs.DC 2026-05 conditional novelty 6.0

    Expert-parallel scaling leaves per-expert routing imbalance flat; mock-token benchmarks overestimate real-text imbalance and fake a batch-size trend; architectures split into data-resilient (MHA, Mamba-2) and persiste...

  5. DisagMoE: Computation-Communication overlapped MoE Training via Disaggregated AF-Pipe Parallelism

    cs.LG 2026-05 unverdicted novelty 6.0

    DisagMoE achieves up to 1.8x faster MoE training by disaggregating attention and FFN layers into disjoint GPU groups with a multi-stage uni-directional pipeline and roofline-based bandwidth balancing.

  6. UniEP: Unified Expert-Parallel MoE MegaKernel for LLM Training

    cs.DC 2026-04 unverdicted novelty 5.0

    UniEP fuses MoE communication and computation into unified MegaKernels with deterministic token ordering, delivering 1.03x-1.38x speedups over prior work while preserving training accuracy.