Mixture-of-Modules: Reinventing Transformers as Dynamic Assemblies of Modules

• URL: http://arxiv.org/abs/2407.06677v1
• Date: Tue, 9 Jul 2024 08:50:18 GMT
• Title: Mixture-of-Modules: Reinventing Transformers as Dynamic Assemblies of Modules
• Authors: Zhuocheng Gong, Ang Lv, Jian Guan, Junxi Yan, Wei Wu, Huishuai Zhang, Minlie Huang, Dongyan Zhao, Rui Yan,
• Abstract summary: We propose a novel architecture dubbed mixture-of-modules (MoM) MoM is motivated by an intuition that any layer, regardless of its position, can be used to compute a token. We show that MoM provides not only a unified framework for Transformers but also a flexible and learnable approach for reducing redundancy.
• Score: 96.21649779507831
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Is it always necessary to compute tokens from shallow to deep layers in Transformers? The continued success of vanilla Transformers and their variants suggests an undoubted "yes". In this work, however, we attempt to break the depth-ordered convention by proposing a novel architecture dubbed mixture-of-modules (MoM), which is motivated by an intuition that any layer, regardless of its position, can be used to compute a token as long as it possesses the needed processing capabilities. The construction of MoM starts from a finite set of modules defined by multi-head attention and feed-forward networks, each distinguished by its unique parameterization. Two routers then iteratively select attention modules and feed-forward modules from the set to process a token. The selection dynamically expands the computation graph in the forward pass of the token, culminating in an assembly of modules. We show that MoM provides not only a unified framework for Transformers and their numerous variants but also a flexible and learnable approach for reducing redundancy in Transformer parameterization. We pre-train various MoMs using OpenWebText. Empirical results demonstrate that MoMs, of different parameter counts, consistently outperform vanilla transformers on both GLUE and XSUM benchmarks. More interestingly, with a fixed parameter budget, MoM-large enables an over 38% increase in depth for computation graphs compared to GPT-2-large, resulting in absolute gains of 1.4 on GLUE and 1 on XSUM. On the other hand, MoM-large also enables an over 60% reduction in depth while involving more modules per layer, yielding a 16% reduction in TFLOPs and a 43% decrease in memory usage compared to GPT-2-large, while maintaining comparable performance.

Related papers

• MoDeGPT: Modular Decomposition for Large Language Model Compression [59.361006801465344]
  This paper introduces textbfModular bfDecomposition (MoDeGPT), a novel structured compression framework. MoDeGPT partitions the Transformer block into modules comprised of matrix pairs and reduces the hidden dimensions. Our experiments show MoDeGPT, without backward propagation, matches or surpasses previous structured compression methods.
  arXiv  Detail & Related papers  (2024-08-19T01:30:14Z)
• HAFormer: Unleashing the Power of Hierarchy-Aware Features for Lightweight Semantic Segmentation [11.334990474402915]
  We introduce HAFormer, a model that combines the hierarchical features extraction ability of CNNs with the global dependency modeling capability of Transformers. HAFormer achieves high performance with minimal computational overhead and compact model size.
  arXiv  Detail & Related papers  (2024-07-10T07:53:24Z)
• MoEUT: Mixture-of-Experts Universal Transformers [75.96744719516813]
  Universal Transformers (UTs) have advantages over standard Transformers in learning compositional generalizations. Layer-sharing drastically reduces the parameter count compared to the non-shared model with the same dimensionality. No previous work has succeeded in proposing a shared-layer Transformer design that is competitive in parameter count-dominated tasks such as language modeling.
  arXiv  Detail & Related papers  (2024-05-25T03:24:32Z)
• SwitchHead: Accelerating Transformers with Mixture-of-Experts Attention [39.09650673080772]
  SwitchHead is an effective MoE method for the attention layer. Our novel MoE mechanism allows SwitchHead to compute up to 8 times fewer attention matrices than the standard Transformer. SwitchHead matches the perplexity of standard models with only 44% compute and 27% memory usage.
  arXiv  Detail & Related papers  (2023-12-13T09:00:21Z)
• Approximating Two-Layer Feedforward Networks for Efficient Transformers [15.793406740545024]
  We present a general framework that unifies various methods to approximate two-layer NNs, including product-key memories (PKMs) We show that our MoEs are competitive with the dense Transformer-XL on both the WikiText-103 and enwiki8 datasets at two different scales. This demonstrates that MoEs are relevant not only to extremely large LMs but also to any-scale resource-efficient LMs.
  arXiv  Detail & Related papers  (2023-10-16T21:23:16Z)
• Isomer: Isomerous Transformer for Zero-shot Video Object Segmentation [59.91357714415056]
  We propose two Transformer variants: Context-Sharing Transformer (CST) and Semantic Gathering-Scattering Transformer (S GST) CST learns the global-shared contextual information within image frames with a lightweight computation; S GST models the semantic correlation separately for the foreground and background. Compared with the baseline that uses vanilla Transformers for multi-stage fusion, ours significantly increase the speed by 13 times and achieves new state-of-the-art ZVOS performance.
  arXiv  Detail & Related papers  (2023-08-13T06:12:00Z)
• AutoMoE: Heterogeneous Mixture-of-Experts with Adaptive Computation for Efficient Neural Machine Translation [104.0979785739202]
  Mixture-of-Expert (MoE) models have obtained state-of-the-art performance in Neural Machine Translation (NMT) tasks. Existing MoE models mostly consider a homogeneous design where the same number of experts of the same size are placed uniformly throughout the network. We develop AutoMoE -- a framework for designing heterogeneous MoE's under computational constraints.
  arXiv  Detail & Related papers  (2022-10-14T05:32:17Z)
• BatchFormerV2: Exploring Sample Relationships for Dense Representation Learning [88.82371069668147]
  BatchFormerV2 is a more general batch Transformer module, which enables exploring sample relationships for dense representation learning. BatchFormerV2 consistently improves current DETR-based detection methods by over 1.3%.
  arXiv  Detail & Related papers  (2022-04-04T05:53:42Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.