CORTEXA
← Browse
arxivcs.DCcs.AI2026-07-02

Mixture-of-Parallelisms: Towards Memory-Efficient Training Stack for Mixture-of-Experts Models

Xuan-Phi Nguyen, Shrey Pandit, Yiran Zhao, Semih Yavuz, Silvio Savarese, Shafiq Joty

This paper showcases a memory-efficient training stack for Mixture-of-Experts (MoE) models. It is a training paradigm that combines and specializes various existing and novel parallelism techniques at different layers and stages of the Mixture-of-Experts (MoE) model training pipeline. It leverages these techniques to achieve maximal efficiency given the physical constraints of CPU, CPU memory, GPU HBM memory, and the CPU-GPU, GPU-GPU, and node-node communication bandwidth of the GPU cluster. It also contains a novel strategy for the optimizer step to achieve high throughput and memory efficiency, enabling practitioners to conduct lossless pre-training/fine-tuning of trillion-parameter scale models, at a million context length, with just under 12 8x H200 GPU nodes, with state-of-the-art throughput and memory efficiency. In our experiments, MoP delivers 4.7x--8.2x higher per-GPU throughput than a strongly-tuned FSDP2 baseline (with the gap widening at larger scale) and sustains training at context lengths up to 1M tokens, where the baseline runs out of memory beyond 64--128K.

View free PDFSource page

Related papers

arxivcs.DCcs.AI2026-07-21

Fine-grained Computation-Communication Overlap via Tile-level Signaling and Scheduling for Mixture-of-Experts

Minyu Cui, Anna Wingkvist, Morgan Ericsson

Mixture-of-Experts (MoE) architectures increase model capacity without proportionally increasing computation cost and have become a key building block for scaling large language models (LLMs) to trillion-parameter regimes. Efficient deployment of these MoE models relies on distri…

View free PDFSource page
arxivcs.CLcs.AIcs.DC2026-07-14

Less Experts, Faster Decoding: Cost-Aware Speculative Decoding for Mixture-of-Experts

Jincheng Xie, Runheng Liu, Heyan Huang, Yawen Ling, Hanbin Dai, Yu Zheng, et al.

Sparse Mixture-of-Experts (MoE) models have become an important approach for scaling Large Language Models (LLMs), but their inference efficiency depends strongly on expert activation patterns. Speculative decoding (SD) accelerates autoregressive generation by verifying multiple…

View free PDFSource page
arxivcs.ARcs.AIcs.DC2026-07-19

ThAME: 3D Memory-Enabled Heterogeneous Accelerator for LLM Mixture of Experts

Pratyush Dhingra, Pramit Kumar Pal, Janardhan Rao Doppa, Partha Pratim Pande

Mixture of Experts (MoE) architectures have emerged as a dominant paradigm for scaling Large Language Models (LLMs). However, MoE inference on conventional hardware is constrained by three fundamental bottlenecks. These encompass the massive memory bandwidth required to fetch non…

View free PDFSource page
arxivcs.DCcs.AIcs.NI2026-07-07

UBEP: Re-architecting Expert Parallelism Communication Library for Production Superpods

Yipeng Liu, Chang Liu, Si Shen, Jiaqi Zheng, Mingfan Li, Yuyang Yang, et al.

The deployment of Mixture-of-Experts (MoE) models on production high-bandwidth superpods, such as NVIDIA's NVL72/576 and Huawei's CloudMatrix384, introduces critical challenges beyond raw interconnect bandwidth. While these systems provide unified global address spaces and high-b…

View free PDFSource page
arxivcs.DCcs.AIcs.SE2026-07-17

JoyNexus: Service-Oriented Multi-Tenant Post-Training for VLA Models

Haoran Sun, Wentao Zhang, Junyang Hua, Hedan Yang, Yongjian Guo, Yifei Zhang, et al.

The post-training of Vision-Language-Action (VLA) models is essential due to the diversity of simulators, robot embodiments, and task objectives. Existing compute services, whether offered as direct accelerator rental or batch-workload submission, typically allocate an exclusive…

View free PDFSource page
arxivcs.ARcs.AIcs.CLcs.DCcs.LGcs.PF2026-07-21

BaseRT: Advancing Best-in-Class LLM Inference with Apple M5 Neural Accelerators

Fabian Waschkowski, Prabod Rathnayaka, Lukas Wesemann

Apple's M5 generation introduces a redesigned GPU architecture in which every core carries a dedicated Neural Accelerator: on-die matrix units exposed through the Metal~4 tensor API. We show that BaseRT, our native Metal inference runtime for large language models on Apple Silico…

View free PDFSource page