A Hybrid Scale-Up and Scale-Out Approach for Performance and Energy Efficiency Optimization in Systolic Array Accelerators.
針對收縮陣列加速器的性能和能量效率優化的混合擴展與擴展方法。
Micromachines (Basel) 2025-03-27
Efficient scaling of large language models with mixture of experts and 3D analog in-memory computing.
大型語言模型的高效擴展:專家混合與3D類比內存計算。
Nat Comput Sci 2025-01-08
2-D Transformer: Extending Large Language Models to Long-Context With Few Memory.
2-D Transformer:擴展大型語言模型以應對長上下文與少量記憶。
IEEE Trans Neural Netw Learn Syst 2025-03-21
Near-Sensor Edge Computing System Enabled by a CMOS Compatible Photonic Integrated Circuit Platform Using Bilayer AlN/Si Waveguides.
基於雙層 AlN/Si 波導、CMOS 相容光子積體電路平台的近感測邊緣運算系統
Nanomicro Lett 2025-05-19