Momentum Transformer: Closing the Performance Gap Between Self-attention and Its Linearization
원문 초록
Transformers have achieved remarkable success in sequence modeling and beyond but suffer from quadratic computational and memory complexities with respect to the length of the input sequence. Leveraging techniques include sparse and linear attention and hashing tricks; efficient transformers have been proposed to reduce the quadratic complexity of transformers but significantly degrade the accuracy. In response, we first interpret the linear attention and residual connections in computing the attention map as gradient descent steps. We then introduce momentum into these components and propose the \emph{momentum transformer}, which utilizes momentum to improve the accuracy of linear transformers while maintaining linear memory and computational complexities. Furthermore, we develop an adaptive strategy to compute the momentum value for our model based on the optimal momentum for quadratic optimization. This adaptive momentum eliminates the need to search for the optimal momentum value and further enhances the performance of the momentum transformer. A range of experiments on both autoregressive and non-autoregressive tasks, including image generation and machine translation, demonstrate that the momentum transformer outperforms popular linear transformers in training efficiency and accuracy.
함께 살펴볼 논문
공통 태그를 근거로 연결했습니다. 실제 인용 관계나 논문의 우수성을 뜻하지 않습니다.
MFGDiffusion: Mask-Guided Smoke Synthesis for Enhanced Forest Fire Detection
Implementing Adaptations for Vision AutoRegressive Model
CharaConsist: Fine-Grained Consistent Character Generation
MGVQ: Could VQ-VAE Beat VAE? A Generalizable Tokenizer with Multi-group Quantization
NeoBabel: A Multilingual Open Tower for Visual Generation
Estimating Object Physical Properties from RGB-D Vision and Depth Robot Sensors Using Deep Learning
토론
이 논문과 연결된 포럼 글이 아직 없습니다.