During the deployment of Large Language Models (LLMs), the autoregressive decoding phase on heterogeneous NPU platforms (e.g., Ascend 910B) faces severe memory-bound challenges. This study reveals the ``Model Scaling Paradox'' caused by the static deployment of single-sized models. It also points out the kernel synchronization overhead of fine-grained speculative decoding \cite{leviathan2023fast, chen2023speculative} under NPU computational graph compilation, and the severe limitations of purely relying on micro-level acceleration algorithms like Prompt LookUp Decoding (PLD)

翻译：在大语言模型（LLM）部署过程中，异质NPU平台（如Ascend 910B）上的自回归解码阶段面临严重的内存瓶颈挑战。本研究揭示了单一尺寸模型静态部署所引发的"模型缩放悖论"，并指出了在NPU计算图编译环境下，细粒度投机解码\cite{leviathan2023fast, chen2023speculative}所引入的内核同步开销，以及纯粹依赖诸如提示查找解码（PLD）等微观加速算法的严重局限性。