Samoyeds: Accelerating MoE Models with Structured Sparsity Leveraging Sparse Tensor Cores

• URL: http://arxiv.org/abs/2503.10725v1
• Date: Thu, 13 Mar 2025 10:34:15 GMT
• Title: Samoyeds: Accelerating MoE Models with Structured Sparsity Leveraging Sparse Tensor Cores
• Authors: Chenpeng Wu, Qiqi Gu, Heng Shi, Jianguo Yao, Haibing Guan,
• Abstract summary: This paper presents Samoyeds, an innovative acceleration system for MoE LLMs utilizing Sparse Cores (SpTCs)<n>It introduces a bespoke sparse data format tailored for MoE computation and develops a specialized sparse-sparse matrix multiplication kernel.<n> Evaluations show that Samoyeds outperforms SOTA works by up to 1.99$times$ at the kernel level and 1.58$times$ at the model level.
• Score: 13.089178890203652
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: The escalating size of Mixture-of-Experts (MoE) based Large Language Models (LLMs) presents significant computational and memory challenges, necessitating innovative solutions to enhance efficiency without compromising model accuracy. Structured sparsity emerges as a compelling strategy to address these challenges by leveraging the emerging sparse computing hardware. Prior works mainly focus on the sparsity in model parameters, neglecting the inherent sparse patterns in activations. This oversight can lead to additional computational costs associated with activations, potentially resulting in suboptimal performance. This paper presents Samoyeds, an innovative acceleration system for MoE LLMs utilizing Sparse Tensor Cores (SpTCs). Samoyeds is the first to apply sparsity simultaneously to both activations and model parameters. It introduces a bespoke sparse data format tailored for MoE computation and develops a specialized sparse-sparse matrix multiplication kernel. Furthermore, Samoyeds incorporates systematic optimizations specifically designed for the execution of dual-side structured sparse MoE LLMs on SpTCs, further enhancing system performance. Evaluations show that Samoyeds outperforms SOTA works by up to 1.99$\times$ at the kernel level and 1.58$\times$ at the model level. Moreover, it enhances memory efficiency, increasing maximum supported batch sizes by 4.41$\times$ on average. Additionally, Samoyeds surpasses existing SOTA structured sparse solutions in both model accuracy and hardware portability.

Related papers

• Every FLOP Counts: Scaling a 300B Mixture-of-Experts LING LLM without Premium GPUs [96.68469559192846]
  We present two differently sized MoE large language models (LLMs) Ling-Lite contains 16.8 billion parameters with 2.75 billion activated parameters, while Ling-Plus boasts 290 billion parameters with 28.8 billion activated parameters. We propose innovative methods for (1) optimization of model architecture and training processes, (2) refinement of training anomaly handling, and (3) enhancement of model evaluation efficiency.
  arXiv  Detail & Related papers  (2025-03-07T04:43:39Z)
• AutoHete: An Automatic and Efficient Heterogeneous Training System for LLMs [68.99086112477565]
  Transformer-based large language models (LLMs) have demonstrated exceptional capabilities in sequence modeling and text generation. Existing heterogeneous training methods significantly expand the scale of trainable models but introduce substantial communication overheads and CPU workloads. We propose AutoHete, an automatic and efficient heterogeneous training system compatible with both single- GPU and multi- GPU environments.
  arXiv  Detail & Related papers  (2025-02-27T14:46:22Z)
• DSMoE: Matrix-Partitioned Experts with Dynamic Routing for Computation-Efficient Dense LLMs [70.91804882618243]
  This paper proposes DSMoE, a novel approach that achieves sparsification by partitioning pre-trained FFN layers into computational blocks.<n>We implement adaptive expert routing using sigmoid activation and straight-through estimators, enabling tokens to flexibly access different aspects of model knowledge.<n>Experiments on LLaMA models demonstrate that under equivalent computational constraints, DSMoE achieves superior performance compared to existing pruning and MoE approaches.
  arXiv  Detail & Related papers  (2025-02-18T02:37:26Z)
• Mamba-Shedder: Post-Transformer Compression for Efficient Selective Structured State Space Models [1.8434042562191815]
  This paper explores the compression of SSM-based models, particularly Mamba and its hybrids.<n>We study the sensitivity of these models to the removal of selected components at different granularities to reduce the model size and computational overhead, thus improving their efficiency while maintaining accuracy.<n>The proposed solutions, collectively referred to as Mamba-Shedder, achieve a speedup of up to 1.4x during inference, demonstrating that model efficiency can be improved by eliminating several redundancies with minimal impact on the overall model performance.
  arXiv  Detail & Related papers  (2025-01-28T17:22:01Z)
• MoE-CAP: Benchmarking Cost, Accuracy and Performance of Sparse Mixture-of-Experts Systems [26.493762260392284]
  MoE-CAP is a benchmarking method for evaluating sparse MoE systems.<n>Its key innovation is a sparsity-aware CAP analysis model, the first to integrate cost, performance, and accuracy metrics into a single diagram.
  arXiv  Detail & Related papers  (2024-12-10T00:19:28Z)
• Structuring a Training Strategy to Robustify Perception Models with Realistic Image Augmentations [1.5723316845301678]
  This report introduces a novel methodology for training with augmentations to enhance model robustness and performance in such conditions. We present a comprehensive framework that includes identifying weak spots in Machine Learning models, selecting suitable augmentations, and devising effective training strategies. Experimental results demonstrate improvements in model performance, as measured by commonly used metrics such as mean Average Precision (mAP) and mean Intersection over Union (mIoU) on open-source object detection and semantic segmentation models and datasets.
  arXiv  Detail & Related papers  (2024-08-30T14:15:48Z)
• Fine-Tuning and Deploying Large Language Models Over Edges: Issues and Approaches [64.42735183056062]
  Large language models (LLMs) have transitioned from specialized models to versatile foundation models. LLMs exhibit impressive zero-shot ability, however, require fine-tuning on local datasets and significant resources for deployment.
  arXiv  Detail & Related papers  (2024-08-20T09:42:17Z)
• Pruning Large Language Models with Semi-Structural Adaptive Sparse Training [17.381160429641316]
  Adaptive Sparse Trainer (AST) is a novel and efficient retraining framework tailored for semi-structured sparse models.<n>AST reduces the perplexity and zero-shot accuracy gap between dense and 2:4 semi-structured sparse models to 0.6 and 1.16%, respectively.
  arXiv  Detail & Related papers  (2024-07-30T06:33:44Z)
• Rethinking Pruning for Vision-Language Models: Strategies for Effective Sparsity and Performance Restoration [41.714059733079374]
  Vision-Language Models (VLMs) integrate information from multiple modalities and have shown remarkable success across various tasks. This study addresses two key questions: how to distribute sparsity across different modality-specific models, and how to restore the performance of pruned sparse VLMs.
  arXiv  Detail & Related papers  (2024-04-03T03:27:01Z)
• XMoE: Sparse Models with Fine-grained and Adaptive Expert Selection [30.687511115573038]
  tool is a novel MoE designed to enhance both the efficacy and efficiency of sparse MoE models. tool can enhance model performance while decreasing the computation load at MoE layers by over 50% without sacrificing performance.
  arXiv  Detail & Related papers  (2024-02-27T08:18:02Z)
• Sparse Modular Activation for Efficient Sequence Modeling [94.11125833685583]
  Recent models combining Linear State Space Models with self-attention mechanisms have demonstrated impressive results across a range of sequence modeling tasks. Current approaches apply attention modules statically and uniformly to all elements in the input sequences, leading to sub-optimal quality-efficiency trade-offs. We introduce Sparse Modular Activation (SMA), a general mechanism enabling neural networks to sparsely activate sub-modules for sequence elements in a differentiable manner.
  arXiv  Detail & Related papers  (2023-06-19T23:10:02Z)
• Scalable and Efficient MoE Training for Multitask Multilingual Models [55.987536562357086]
  We develop a system capable of scaling MoE models efficiently to trillions of parameters. We also present new training methods to improve MoE sample efficiency and leverage expert pruning strategy to improve time efficiency. A model trained with 10 billion parameters on 50 languages can achieve state-of-the-art performance in Machine Translation (MT) and multilingual natural language generation tasks.
  arXiv  Detail & Related papers  (2021-09-22T00:57:46Z)
• Optimization-driven Machine Learning for Intelligent Reflecting Surfaces Assisted Wireless Networks [82.33619654835348]
  Intelligent surface (IRS) has been employed to reshape the wireless channels by controlling individual scattering elements' phase shifts. Due to the large size of scattering elements, the passive beamforming is typically challenged by the high computational complexity. In this article, we focus on machine learning (ML) approaches for performance in IRS-assisted wireless networks.
  arXiv  Detail & Related papers  (2020-08-29T08:39:43Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.