发表机构
Princeton University; Google Deepmind; PPPL（普林斯顿大学; 谷歌DeepMind; 普林斯顿等离子体物理实验室）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
本文针对核聚变控制中机器学习推理速度问题，基于DIII-D托卡马克经验开发基准，比较CPU与GPU后端性能，指出大模型需GPU且部署后端应与模型及控制预算协同选择。
AI 中文摘要
机器学习模型越来越多地用于核聚变的反馈控制回路中，其中推理速度和可预测的时序至关重要。我们总结了在DIII-D托卡马克上部署用于控制的模型的经验教训，并开发了一个基准测试，用于比较来自聚变控制和诊断管道的十个神经网络及模型组件的推理后端。对于参数超过五百万的模型，CPU后端需要数十到数千毫秒，而GPU推理则明显更快，这表明CPU导向的开发在控制方面存在上限。这些结果说明了为什么部署后端必须与模型及其控制周期预算一起选择。
英文摘要
Machine learning models are increasingly used in feedback control loops for nuclear fusion, where inference speed and predictable timing are critical. We summarize lessons from models deployed for control on the DIII-D tokamak and develop a benchmark to compare inference backends across ten neural networks and model components from fusion control and diagnostic pipelines. For models greater than five million parameters, the CPU backends take tens to thousands of milliseconds, while GPU inference is substantially faster, suggesting an upper limit on CPU-oriented development for control. These results show why the deployment backend must be selected together with the model and its control-cycle budget.
