Duplex: A Device for Large Language Models with Mixture of Experts, Grouped Query Attention, and Continuous Batching

• URL: http://arxiv.org/abs/2409.01141v1
• Date: Mon, 2 Sep 2024 10:21:21 GMT
• Title: Duplex: A Device for Large Language Models with Mixture of Experts, Grouped Query Attention, and Continuous Batching
• Authors: Sungmin Yun, Kwanhee Kyung, Juhwan Cho, Jaewan Choi, Jongmin Kim, Byeongho Kim, Sukhan Lee, Kyomin Sohn, Jung Ho Ahn,
• Abstract summary: We observe that conventional computing devices have limitations when processing the MoE and attention layers. To address these challenges, we propose xPU tailored for low-Op/B and LogicPIM tailored for low-Op/B operations.
• Score: 2.863328705885669
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Large language models (LLMs) have emerged due to their capability to generate high-quality content across diverse contexts. To reduce their explosively increasing demands for computing resources, a mixture of experts (MoE) has emerged. The MoE layer enables exploiting a huge number of parameters with less computation. Applying state-of-the-art continuous batching increases throughput; however, it leads to frequent DRAM access in the MoE and attention layers. We observe that conventional computing devices have limitations when processing the MoE and attention layers, which dominate the total execution time and exhibit low arithmetic intensity (Op/B). Processing MoE layers only with devices targeting low-Op/B such as processing-in-memory (PIM) architectures is challenging due to the fluctuating Op/B in the MoE layer caused by continuous batching. To address these challenges, we propose Duplex, which comprises xPU tailored for high-Op/B and Logic-PIM to effectively perform low-Op/B operation within a single device. Duplex selects the most suitable processor based on the Op/B of each layer within LLMs. As the Op/B of the MoE layer is at least 1 and that of the attention layer has a value of 4-8 for grouped query attention, prior PIM architectures are not efficient, which place processing units inside DRAM dies and only target extremely low-Op/B (under one) operations. Based on recent trends, Logic-PIM adds more through-silicon vias (TSVs) to enable high-bandwidth communication between the DRAM die and the logic die and place powerful processing units on the logic die, which is best suited for handling low-Op/B operations ranging from few to a few dozens. To maximally utilize the xPU and Logic-PIM, we propose expert and attention co-processing.

Related papers

• The New LLM Bottleneck: A Systems Perspective on Latent Attention and Mixture-of-Experts [5.10053312713569]
  This paper argues that recent architectural shifts, namely Multi-head Latent Attention (MLA) and Mixture-of-Experts (MoE), challenge the premise of specialized attention hardware.<n>The central challenge for next-generation Transformers is no longer accelerating single memory-bound layer.<n>Instead, the focus must shift to designing balanced systems with sufficient compute memory capacity, memory bandwidth, and high-bandwidth interconnects to manage the diverse demands of large-scale models.
  arXiv  Detail & Related papers  (2025-07-21T10:18:33Z)
• Inter2Former: Dynamic Hybrid Attention for Efficient High-Precision Interactive [58.0729162588429]
  Interactive segmentation improves annotation efficiency by segmenting target regions from user prompts.<n>Current approaches face a critical trade-off: dense-token methods achieve superior accuracy but suffer from prohibitively slow processing on CPU devices.<n>We propose Inter2Former to address this challenge by optimizing computation allocation in dense-token processing.
  arXiv  Detail & Related papers  (2025-07-13T12:33:37Z)
• MEM1: Learning to Synergize Memory and Reasoning for Efficient Long-Horizon Agents [84.62985963113245]
  We introduce MEM1, an end-to-end reinforcement learning framework that enables agents to operate with constant memory across long multi-turn tasks.<n>At each turn, MEM1 updates a compact shared internal state that jointly supports memory consolidation and reasoning.<n>We show that MEM1-7B improves performance by 3.5x while reducing memory usage by 3.7x compared to Qwen2.5-14B-Instruct on a 16-objective multi-hop QA task.
  arXiv  Detail & Related papers  (2025-06-18T19:44:46Z)
• Pangu Embedded: An Efficient Dual-system LLM Reasoner with Metacognition [95.54406667705999]
  Pangu Embedded is an efficient Large Language Model (LLM) reasoner developed on Ascend Neural Processing Units (NPUs)<n>It addresses the significant computational costs and inference latency challenges prevalent in existing reasoning-optimized LLMs.<n>It delivers rapid responses and state-of-the-art reasoning quality within a single, unified model architecture.
  arXiv  Detail & Related papers  (2025-05-28T14:03:02Z)
• D$^{2}$MoE: Dual Routing and Dynamic Scheduling for Efficient On-Device MoE-based LLM Serving [14.607254882119507]
  Combination of experts (MoE) model is a sparse variant of large language models (LLMs) Despite its benefits, MoE is still too expensive to deploy on resource-constrained edge devices. We propose D$2$MoE, an algorithm-system co-design framework that matches diverse task requirements by dynamically allocating the most proper bit-width to each expert.
  arXiv  Detail & Related papers  (2025-04-17T05:37:35Z)
• MegaScale-Infer: Serving Mixture-of-Experts at Scale with Disaggregated Expert Parallelism [26.923312725688735]
  Mixture-of-Experts (MoE) showcases tremendous potential to scale large language models (LLMs) with enhanced performance and reduced computational complexity. We present MegaScale-Infer, an efficient and cost-effective system for serving large-scale MoE models.
  arXiv  Detail & Related papers  (2025-04-03T04:20:44Z)
• Optimizing Multi-DNN Inference on Mobile Devices through Heterogeneous Processor Co-Execution [39.033040759452504]
  Deep Neural Networks (DNNs) are increasingly deployed across diverse industries, driving demand for mobile device support. Existing mobile inference frameworks often rely on a single processor per model, limiting hardware utilization and causing suboptimal performance and energy efficiency. We propose an Advanced Multi-DNN Model Scheduling (ADMS) strategy for optimizing multi-DNN inference on mobile heterogeneous processors.
  arXiv  Detail & Related papers  (2025-03-27T03:03:09Z)
• PowerAttention: Exponentially Scaling of Receptive Fields for Effective Sparse Attention [73.26995918610669]
  Large Language Models (LLMs) face efficiency bottlenecks due to the quadratic complexity of the attention mechanism when processing long contexts. We introduce PowerAttention, a novel sparse attention design that facilitates effective and complete context extension. Experiments demonstrate that PowerAttention outperforms existing static sparse attention methods by $5sim 40%$.
  arXiv  Detail & Related papers  (2025-03-05T15:24:11Z)
• DeeR-VLA: Dynamic Inference of Multimodal Large Language Models for Efficient Robot Execution [114.61347672265076]
  Development of MLLMs for real-world robots is challenging due to the typically limited computation and memory capacities available on robotic platforms. We propose a Dynamic Early-Exit Framework for Robotic Vision-Language-Action Model (DeeR) that automatically adjusts the size of the activated MLLM. DeeR demonstrates significant reductions in computational costs of LLM by 5.2-6.5x and GPU memory of LLM by 2-6x without compromising performance.
  arXiv  Detail & Related papers  (2024-11-04T18:26:08Z)
• DynaSplit: A Hardware-Software Co-Design Framework for Energy-Aware Inference on Edge [40.96858640950632]
  We propose DynaSplit, a framework that dynamically configures parameters across both software and hardware. We evaluate DynaSplit using popular pre-trained NNs on a real-world testbed. Results show a reduction in energy consumption up to 72% compared to cloud-only computation.
  arXiv  Detail & Related papers  (2024-10-31T12:44:07Z)
• EPS-MoE: Expert Pipeline Scheduler for Cost-Efficient MoE Inference [49.94169109038806]
  This paper introduces EPS-MoE, a novel expert pipeline scheduler for MoE. Our results demonstrate an average 21% improvement in prefill throughput over existing parallel inference methods.
  arXiv  Detail & Related papers  (2024-10-16T05:17:49Z)
• OPIMA: Optical Processing-In-Memory for Convolutional Neural Network Acceleration [5.0389804644646174]
  We introduce OPIMA, a processing-in-memory (PIM)-based machine learning accelerator. PIM struggles to achieve high throughput and energy efficiency due to internal data movement bottlenecks. We show that OPIMA can achieve 2.98x higher throughput and 137x better energy efficiency than the best-known prior work.
  arXiv  Detail & Related papers  (2024-07-11T06:12:04Z)
• Efficient and Economic Large Language Model Inference with Attention Offloading [11.698376311689456]
  Transformer-based large language models (LLMs) exhibit impressive performance in generative tasks but introduce significant challenges in real-world serving. This mismatch arises from the autoregressive nature of LLMs, where the generation phase comprises operators with varying resource demands. To enhance the efficiency and cost-effectiveness of LLM serving, we introduce the concept of attention offloading.
  arXiv  Detail & Related papers  (2024-05-03T02:15:15Z)
• HiRE: High Recall Approximate Top-$k$ Estimation for Efficient LLM Inference [68.59839755875252]
  HiRE comprises of two novel components: (i) a compression scheme to cheaply predict top-$k$ rows/columns with high recall, followed by full computation restricted to the predicted subset, and (ii) DA-TOP-$k$: an efficient multi-device approximate top-$k$ operator. We demonstrate that on a one billion parameter model, HiRE applied to both the softmax as well as feedforward layers, achieves almost matching pretraining and downstream accuracy, and speeds up inference latency by $1.47times$ on a single TPUv5e device.
  arXiv  Detail & Related papers  (2024-02-14T18:04:36Z)
• Distributed Inference and Fine-tuning of Large Language Models Over The Internet [91.00270820533272]
  Large language models (LLMs) are useful in many NLP tasks and become more capable with size. These models require high-end hardware, making them inaccessible to most researchers. We develop fault-tolerant inference algorithms and load-balancing protocols that automatically assign devices to maximize the total system throughput.
  arXiv  Detail & Related papers  (2023-12-13T18:52:49Z)
• Edge-MoE: Memory-Efficient Multi-Task Vision Transformer Architecture with Task-level Sparsity via Mixture-of-Experts [60.1586169973792]
  M$3$ViT is the latest multi-task ViT model that introduces mixture-of-experts (MoE) MoE achieves better accuracy and over 80% reduction computation but leaves challenges for efficient deployment on FPGA. Our work, dubbed Edge-MoE, solves the challenges to introduce the first end-to-end FPGA accelerator for multi-task ViT with a collection of architectural innovations.
  arXiv  Detail & Related papers  (2023-05-30T02:24:03Z)
• M$^3$ViT: Mixture-of-Experts Vision Transformer for Efficient Multi-task Learning with Model-Accelerator Co-design [95.41238363769892]
  Multi-task learning (MTL) encapsulates multiple learned tasks in a single model and often lets those tasks learn better jointly. Current MTL regimes have to activate nearly the entire model even to just execute a single task. We present a model-accelerator co-design framework to enable efficient on-device MTL.
  arXiv  Detail & Related papers  (2022-10-26T15:40:24Z)

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.