使用复用技术简化变形金刚网络以用于语音识别的 Sim-T（英文版）

英文标题：Sim-T: Simplify the Transformer Network by Multiplexing Technique for Speech Recognition中文摘要：本文提出了一种新型轻量级模型 Sim-T，通过分别运用模块权重多路复用和注意力分数多路复用两种技术，实现了对 Transformer 模型的有效压缩，并验证了其在语音识别任务上的高效性和优越性。英文摘要：In recent years, a great deal of attention has been paid to the Transformernetwork for speech recognition tasks due to its excellent model performance.However, the Transformer network always involves heavy computation and largenumber of parameters, causing serious deployment problems in devices wit