English

HyperParallel-MoE: Multi-Core Interleaved Scheduling for Fast MoE Training on Ascend NPUs

Distributed, Parallel, and Cluster Computing 2026-05-25 v1

Abstract

Modern Mixture-of-Experts (MoE) models increasingly rely on large-scale AI accelerator clusters for efficient training. Ascend NPUs expose heterogeneous on-chip compute resources, including matrix-oriented AIC units and vector-oriented AIV units with explicit cross-queue synchronization support. However, existing training frameworks largely execute MoE operators in a serialized kernel-by-kernel manner, leaving substantial heterogeneous parallelism underutilized. This paper presents HyperParallel-MoE, a compilation and scheduling framework for MoE training on Ascend NPUs. HyperParallel-MoE transforms operator-level MoE execution into a statically scheduled tile-level heterogeneous taskflow spanning AIC and AIV resources. It introduces AIV-driven one-sided communication to eliminate host-side collective synchronization, dependency-preserving tile task generation to unify communication and computation under a common task abstraction, and event-driven static scheduling to coordinate cross-queue execution with low runtime overhead. HyperParallel-MoE further executes the compiled taskflow within a unified runtime that concurrently drives AIC and AIV workers inside a single kernel launch, enabling fine-grained overlap among communication, matrix computation, and vector computation while preserving existing optimized operators. We implement HyperParallel-MoE in the MindSpore and MindFormers stack and evaluate it using DeepSeek-style MoE models on Ascend A3 clusters. Across multiple expert-parallel configurations, HyperParallel-MoE reduces Dispatch-to-Combine MoE-FFN latency by up to 1.58x, demonstrating that tile-level heterogeneous scheduling can substantially improve MoE training efficiency on modern NPUs.

Keywords

Cite

@article{arxiv.2605.23764,
  title  = {HyperParallel-MoE: Multi-Core Interleaved Scheduling for Fast MoE Training on Ascend NPUs},
  author = {Zewen Jin and Congkun Ai and Guangpeng Zhang and Hanbo Zhang and Haoran Wang and Shihan Xiao and Da Lei and Xuefeng Jin and Teng Su and Cheng Li},
  journal= {arXiv preprint arXiv:2605.23764},
  year   = {2026}
}