Efficient Diffusion Transformer with Step-wise Dynamic Attention Mediators

• URL: http://arxiv.org/abs/2408.05710v1
• Date: Sun, 11 Aug 2024 07:01:39 GMT
• Title: Efficient Diffusion Transformer with Step-wise Dynamic Attention Mediators
• Authors: Yifan Pu, Zhuofan Xia, Jiayi Guo, Dongchen Han, Qixiu Li, Duo Li, Yuhui Yuan, Ji Li, Yizeng Han, Shiji Song, Gao Huang, Xiu Li,
• Abstract summary: We present a novel diffusion transformer framework incorporating an additional set of mediator tokens to engage with queries and keys separately. Our model initiates the denoising process with a precise, non-ambiguous stage and gradually transitions to a phase enriched with detail. Our method achieves a state-of-the-art FID score of 2.01 when integrated with the recent work SiT.
• Score: 83.48423407316713
• License: http://creativecommons.org/licenses/by-nc-sa/4.0/
• Abstract: This paper identifies significant redundancy in the query-key interactions within self-attention mechanisms of diffusion transformer models, particularly during the early stages of denoising diffusion steps. In response to this observation, we present a novel diffusion transformer framework incorporating an additional set of mediator tokens to engage with queries and keys separately. By modulating the number of mediator tokens during the denoising generation phases, our model initiates the denoising process with a precise, non-ambiguous stage and gradually transitions to a phase enriched with detail. Concurrently, integrating mediator tokens simplifies the attention module's complexity to a linear scale, enhancing the efficiency of global attention processes. Additionally, we propose a time-step dynamic mediator token adjustment mechanism that further decreases the required computational FLOPs for generation, simultaneously facilitating the generation of high-quality images within the constraints of varied inference budgets. Extensive experiments demonstrate that the proposed method can improve the generated image quality while also reducing the inference cost of diffusion transformers. When integrated with the recent work SiT, our method achieves a state-of-the-art FID score of 2.01. The source code is available at https://github.com/LeapLabTHU/Attention-Mediators.

Related papers

• Scaled and Inter-token Relation Enhanced Transformer for Sample-restricted Residential NILM [0.0]
  We propose two novel mechanisms to enhance the attention mechanism of the original transformer to improve performance. The first mechanism reduces the prioritization of intra-token relationships in the token similarity matrix during training, thereby increasing inter-token focus. The second mechanism introduces a learnable temperature tuning for the token similarity matrix, mitigating the over-smoothing problem associated with fixed temperature values.
  arXiv  Detail & Related papers  (2024-10-12T18:58:45Z)
• PointMT: Efficient Point Cloud Analysis with Hybrid MLP-Transformer Architecture [46.266960248570086]
  This study tackles the quadratic complexity of the self-attention mechanism by introducing a complexity local attention mechanism for effective feature aggregation. We also introduce a parameter-free channel temperature adaptation mechanism that adaptively adjusts the attention weight distribution in each channel. We show that PointMT achieves performance comparable to state-of-the-art methods while maintaining an optimal balance between performance and accuracy.
  arXiv  Detail & Related papers  (2024-08-10T10:16:03Z)
• FORA: Fast-Forward Caching in Diffusion Transformer Acceleration [39.51519525071639]
  Diffusion transformers (DiT) have become the de facto choice for generating high-quality images and videos. Fast-FORward CAching (FORA) is designed to accelerate DiT by exploiting the repetitive nature of the diffusion process.
  arXiv  Detail & Related papers  (2024-07-01T16:14:37Z)
• Skip-Layer Attention: Bridging Abstract and Detailed Dependencies in Transformers [56.264673865476986]
  This paper introduces Skip-Layer Attention (SLA) to enhance Transformer models. SLA improves the model's ability to capture dependencies between high-level abstract features and low-level details. Our implementation extends the Transformer's functionality by enabling queries in a given layer to interact with keys and values from both the current layer and one preceding layer.
  arXiv  Detail & Related papers  (2024-06-17T07:24:38Z)
• Towards Better Text-to-Image Generation Alignment via Attention Modulation [16.020834525343997]
  We propose an attribution-focusing mechanism, a training-free phase-wise mechanism by modulation of attention for diffusion model. An object-focused masking scheme and a phase-wise dynamic weight control mechanism are integrated into the cross-attention modules. The experimental results in various alignment scenarios demonstrate that our model attains better image-text alignment with minimal additional computational cost.
  arXiv  Detail & Related papers  (2024-04-22T06:18:37Z)
• Diffusion Models Without Attention [110.5623058129782]
  Diffusion State Space Model (DiffuSSM) is an architecture that supplants attention mechanisms with a more scalable state space model backbone. Our focus on FLOP-efficient architectures in diffusion training marks a significant step forward.
  arXiv  Detail & Related papers  (2023-11-30T05:15:35Z)
• Towards More Accurate Diffusion Model Acceleration with A Timestep Aligner [84.97253871387028]
  A diffusion model, which is formulated to produce an image using thousands of denoising steps, usually suffers from a slow inference speed. We propose a timestep aligner that helps find a more accurate integral direction for a particular interval at the minimum cost. Experiments show that our plug-in design can be trained efficiently and boost the inference performance of various state-of-the-art acceleration methods.
  arXiv  Detail & Related papers  (2023-10-14T02:19:07Z)
• Mutual Information-driven Triple Interaction Network for Efficient Image Dehazing [54.168567276280505]
  We propose a novel Mutual Information-driven Triple interaction Network (MITNet) for image dehazing. The first stage, named amplitude-guided haze removal, aims to recover the amplitude spectrum of the hazy images for haze removal. The second stage, named phase-guided structure refined, devotes to learning the transformation and refinement of the phase spectrum.
  arXiv  Detail & Related papers  (2023-08-14T08:23:58Z)
• Towards Robust Video Instance Segmentation with Temporal-Aware Transformer [12.81807735850422]
  We propose TAFormer to aggregate-aware temporal features in encoder and decoder. TAFormer effectively leverages the spatial and temporal information to obtain context-aware feature representation and outperforms state-of-the-art methods.
  arXiv  Detail & Related papers  (2023-01-20T05:22:16Z)
• CCVS: Context-aware Controllable Video Synthesis [95.22008742695772]
  presentation introduces a self-supervised learning approach to the synthesis of new video clips from old ones. It conditions the synthesis process on contextual information for temporal continuity and ancillary information for fine control.
  arXiv  Detail & Related papers  (2021-07-16T17:57:44Z)

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.