Transformer-based language models spread FLOPs uniformly across input sequences. In this work we demonstrate that transformers can instead learn to dynamically allocate FLOPs (or compute) to specific positions in a sequence, optimising the allocation along the sequence for different layers across the model depth. Our method enforces a total compute budget by capping the number of tokens ($k$) that can participate in the self-attention and MLP computations at a given layer. The tokens to be processed are determined by the network using a top-$k$ routing mechanism. Since $k$ is defined a priori, this simple procedure uses a static computation graph with known tensor sizes, unlike other conditional computation techniques. Nevertheless, since the identities of the $k$ tokens are fluid, this method can expend FLOPs non-uniformly across the time and model depth dimensions. Thus, compute expenditure is entirely predictable in sum total, but dynamic and context-sensitive at the token-level. Not only do models trained in this way learn to dynamically allocate compute, they do so efficiently. These models match baseline performance for equivalent FLOPS and wall-clock times to train, but require a fraction of the FLOPs per forward pass, and can be upwards of 50\% faster to step during post-training sampling.
Evolutionary Soups: Evolving Mixture-of-Experts for Multi-Objective LLM Alignment
Fraunhofer Institute for Applied Information Technology FIT(弗劳恩霍夫应用信息技术研究所FIT);University of Cologne(科隆大学);University of Stuttgart(斯图加特大学);University of Southampton(南安普顿大学);Soochow University(苏州大学);University Hospital of Cologne(科隆大学医院)
VA-MoE: Variables-Adaptive Mixture of Experts for Incremental Weather Forecasting
Hong Kong University of Science and Technology (HKUST)(香港科学与技术大学);Huawei Technologies Ltd.(华为技术有限公司);Zhejiang University(浙江大学);Shanghai AI Laboratory(上海人工智能实验室)
Multi-level Mixture of Experts for Multimodal Entity Linking
School of Computer;Information Technology Shanxi University Taiyuan China;School of Computer Science;Informatics Cardiff University Cardiff UK;ILCC, School of Informatics University of Edinburgh Edinburgh UK;Information Technology Shanxi University;Informatics Cardiff University;ILCC, School of Informatics University of Edinburgh
Mixture of ELM based experts with trainable gating network
Yazd University(亚兹德大学)
混合专家模型方法是一种基于神经网络的集成学习方法,具有很强的提高整体分类精度的能力。该方法基于分而治之的原则,通过选通网络的监督将问题空间划分给多个专家。本文提出了一种基于混合专家模型的集成学习方法--MEETG(mixed of ELM based experts with trainable gating networks),它由多层感知器(MLP)和门网络组成,其中应用基于梯度的学习算法来训练MLP,这是一个迭代且耗时的过程。为了克服这些问题,我们利用极限学习机(ELM)的优点设计了ME的结构,ELM作为单隐层前馈神经网络的学习算法,与其他传统的学习算法相比,具有更快的学习速度和更好的泛化能力。此外,在所提出的方法中,一个可训练的门控网络被应用于根据输入样本动态地聚集专家的输出。在11个基准数据集上的实验结果和统计分析证实了MEETG在分类问题上具有可接受的性能。此外,我们的实验结果表明,该方法在预测稳定性和分类准确性方面优于原始ELM。
Merging Multi-Task Models via Weight-Ensembling Mixture of Experts
National Engineering Research Center for Multimedia Software(国家多媒体软件工程技术研究中心);School of Computer Science, Wuhan University(武汉大学计算机学院);Hubei Luojia Laboratory(湖北珞珈实验室);Sun Yat-sen University(中山大学);JD Explore Academy(京东探索研究院);Mohamed bin Zayed University of Artificial Intelligence(穆罕默德·本·扎耶德人工智能大学);Nanyang Technological University(南洋理工大学)
MoLF: Mixture-of-Latent-Flow for Pan-Cancer Spatial Gene Expression Prediction from Histology
Translational Surgical Oncology, National Center for Tumor Diseases (NCT/UCC) Dresden, Germany;Faculty of Medicine;University Hospital Carl Gustav Carus, Dresden University of Technology;German Cancer Research Center (DKFZ), Heidelberg, Germany
Scaling Continual Learning to 300+ Tasks with Bi-Level Routing Mixture-of-Experts
School of Computing and Data Science, The University of Hong Kong(计算与数据科学学院,香港大学);Center for Embodied Artificial Intelligence and Computer Vision, Shenzhen Loop Area Institute(具身人工智能与计算机视觉中心,深圳环宇区研究所)
PWC-MoE: Privacy-Aware Wireless Collaborative Mixture of Experts
Department of Engineering, King’s College London(伦敦国王学院工程系);Institute for Digital Communication, Friedrich-Alexander-Universität Erlangen-Nürnberg(埃朗根-纽伦堡弗里德里希-亚历山大大学数字通信研究所)
HyperMoE: Towards Better Mixture of Experts via Transferring Among Experts
Beijing University of Posts and Telecommunications(北京邮电大学);Tsinghua University(清华大学);INF Technology(INF科技);Hong Kong University of Science and Technology(香港科技大学)
MixLoRA-DSI: Dynamically Expandable Mixture-of-LoRA Experts for Rehearsal-Free Generative Retrieval over Dynamic Corpora
Department of Data Science & AI, Monash University, Australia(数据科学与人工智能系,墨尔本大学,澳大利亚);Department of Human Centred Computing, Monash University, Australia(以人为中心计算系,墨尔本大学,澳大利亚)
Aioli: A Unified Optimization Framework for Language Model Data Mixing
Department of Computer Science, Stanford University(斯坦福大学计算机科学系);Center for Data Science, New York University(纽约大学数据科学中心);Computer Science Department, New York University(纽约大学计算机科学系);Prescient Design, Genentech(基因泰克预见设计)
AdaMixT: Adaptive Weighted Mixture of Multi-Scale Expert Transformers for Time Series Forecasting
School of Computer Science, Peking University(北京大学计算机科学系);Center for Machine Learning Research, Peking University(北京大学机器学习研究中心);Tsinghua University(清华大学)