TIDE: 一种高效且无损的MoE扩散大语言模型推理方法
TIDE: Efficient and Lossless MoE Diffusion LLM Inference with I/O-aware Expert Offload
AI总结：
本文提出TIDE，一种基于扩散过程块内专家激活时间稳定性的新推理系统，通过引入基于区间的专家刷新策略，优化I/O和CPU计算，实现无损加速，提升推理效率。
摘要：
扩散大语言模型（dLLMs）作为一种具有更好硬件利用和双向上下文能力的替代方案，通过并行块级解码出现。然而，随着dLLMs在混合专家（MoE）架构下不断扩展，其在资源受限设备上的部署仍是一个开放性挑战。现有基于自回归（AR）的方法通常导致严重的I/O开销或显著的计算瓶颈。在本文中，我们提出TIDE，一种新的资源高效的推理系统，利用扩散过程中块内专家激活的时序稳定性。具体来说，我们利用扩散过程中块内专家激活的时序稳定性，并引入基于区间的专家刷新策略，以I/O-aware的方式更新专家位置。为了确保最佳性能，我们将推理调度形式化为一个数学规划问题，求解最优区间以最小化I/O流量和CPU计算。最重要的是，TIDE是一种无损优化，不需要模型训练，为dLLM推理提供
英文摘要：
Diffusion Large Language Models (dLLMs) have emerged as a competitive alternative to autoregressive (AR) models, offering better hardware utilization and bidirectional context through parallel block-level decoding. However, as dLLMs continue to scale up with mixture-of-experts (MoE) architectures, their deployment on resource-constrained devices remains an open challenge. Existing AR-based methods often incur either prohibitive I/O overhead or significant compute bottlenecks. In this work, we propose TIDE, a novel resource-efficient inference system that leverages the temporal stability of expert activations during the diffusion process within the block. Specifically, we leverage the temporal stability of expert activations during the diffusion process within the block and introduce an interval-based expert refresh strategy that updates the expert placement in an I/O-aware fashion. To ensure optimal performance, we formulate the inference scheduling as a mathematical programming problem, solving for the optimal interval that minimizes I/O traffic and CPU computation. Most importantly, TIDE is a lossless optimization that requires no model training, providing a "free lunch" acceleration for dLLM inference. In a single GPU-CPU system, we demonstrate that TIDE achieves up to 1.4$\times$ and 1.5$\times$ throughput improvements over prior baselines on LLaDA2.0-mini and LLaDA2.0-flash models, respectively.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
