EDiT: Efficient Diffusion Transformers with Linear Compressed Attention

• URL: http://arxiv.org/abs/2503.16726v1
• Date: Thu, 20 Mar 2025 21:58:45 GMT
• Title: EDiT: Efficient Diffusion Transformers with Linear Compressed Attention
• Authors: Philipp Becker, Abhinav Mehrotra, Ruchika Chavhan, Malcolm Chadwick, Luca Morreale, Mehdi Noroozi, Alberto Gil Ramos, Sourav Bhattacharya,
• Abstract summary: quadratic scaling properties of the attention in DiTs hinder image generation with higher resolution or on devices with limited resources.<n>We introduce an efficient diffusion transformer (EDiT) to alleviate these efficiency bottlenecks.<n>We demonstrate the effectiveness of the EDiT and MM-EDiT architectures by integrating them into PixArt-Sigma(conventional DiT) and Stable Diffusion 3.5-Medium (MM-DiT)
• Score: 11.36660486878447
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Diffusion Transformers (DiTs) have emerged as a leading architecture for text-to-image synthesis, producing high-quality and photorealistic images. However, the quadratic scaling properties of the attention in DiTs hinder image generation with higher resolution or on devices with limited resources. This work introduces an efficient diffusion transformer (EDiT) to alleviate these efficiency bottlenecks in conventional DiTs and Multimodal DiTs (MM-DiTs). First, we present a novel linear compressed attention method that uses a multi-layer convolutional network to modulate queries with local information while keys and values are spatially aggregated. Second, we formulate a hybrid attention scheme for multi-modal inputs that combines linear attention for image-to-image interactions and standard scaled dot-product attention for interactions involving prompts. Merging these two approaches leads to an expressive, linear-time Multimodal Efficient Diffusion Transformer (MM-EDiT). We demonstrate the effectiveness of the EDiT and MM-EDiT architectures by integrating them into PixArt-Sigma(conventional DiT) and Stable Diffusion 3.5-Medium (MM-DiT), achieving up to 2.2x speedup with comparable image quality after distillation.

Related papers

• DiTFastAttnV2: Head-wise Attention Compression for Multi-Modality Diffusion Transformers [8.697604549478159]
  We introduce DiTFastAttnV2, a post-training compression method designed to accelerate attention in text-to-image generation models. Through an in-depth analysis of MMDiT's attention patterns, we propose head-wise arrow attention and caching mechanisms. With the customized kernel, DiTFastAttnV2 achieves a 68% reduction in attention FLOPs and 1.5x end-to-end speedup on 2K image generation.
  arXiv  Detail & Related papers  (2025-03-28T18:00:12Z)
• FUSE: Label-Free Image-Event Joint Monocular Depth Estimation via Frequency-Decoupled Alignment and Degradation-Robust Fusion [63.87313550399871]
  Image-event joint depth estimation methods leverage complementary modalities for robust perception, yet face challenges in generalizability. We propose Self-supervised Transfer (PST) and FrequencyDe-coupled Fusion module (FreDF) PST establishes cross-modal knowledge transfer through latent space alignment with image foundation models. FreDF explicitly decouples high-frequency edge features from low-frequency structural components, resolving modality-specific frequency mismatches.
  arXiv  Detail & Related papers  (2025-03-25T15:04:53Z)
• DTU-Net: A Multi-Scale Dilated Transformer Network for Nonlinear Hyperspectral Unmixing [16.19039818961399]
  We propose a Dilated Transformer-based unmixing network for nonlinear hyperspectral unmixing.<n>The decoder is designed to accommodate both linear and nonlinear mixing scenarios.<n>Its interpretability is enhanced by explicitly modeling the relationships between endmembers, abundances, and nonlinear coefficients.
  arXiv  Detail & Related papers  (2025-03-05T12:56:33Z)
• Layer- and Timestep-Adaptive Differentiable Token Compression Ratios for Efficient Diffusion Transformers [55.87192133758051]
  Diffusion Transformers (DiTs) have achieved state-of-the-art (SOTA) image generation quality but suffer from high latency and memory inefficiency.<n>We propose DiffRatio-MoD, a dynamic DiT inference framework with differentiable compression ratios.
  arXiv  Detail & Related papers  (2024-12-22T02:04:17Z)
• OminiControl: Minimal and Universal Control for Diffusion Transformer [68.3243031301164]
  We present OminiControl, a novel approach that rethinks how image conditions are integrated into Diffusion Transformer (DiT) architectures.<n>OminiControl addresses these limitations through three key innovations.
  arXiv  Detail & Related papers  (2024-11-22T17:55:15Z)
• FiTv2: Scalable and Improved Flexible Vision Transformer for Diffusion Model [76.84519526283083]
  We present the textbfFlexible Vision Transformer (FiT), a transformer architecture specifically designed for generating images with textitunrestricted resolutions and aspect ratios FiTv2 exhibits $2times$ convergence speed of FiT, when incorporating advanced training-free extrapolation techniques. Comprehensive experiments demonstrate the exceptional performance of FiTv2 across a broad range of resolutions.
  arXiv  Detail & Related papers  (2024-10-17T15:51:49Z)
• Efficient Diffusion Transformer with Step-wise Dynamic Attention Mediators [83.48423407316713]
  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.
  arXiv  Detail & Related papers  (2024-08-11T07:01:39Z)
• MDT-A2G: Exploring Masked Diffusion Transformers for Co-Speech Gesture Generation [44.74056930805525]
  We introduce a novel Masked Diffusion Transformer for co-speech gesture generation, referred to as MDT-A2G. This model employs a mask modeling scheme specifically designed to strengthen temporal relation learning among sequence gestures. Experimental results demonstrate that MDT-A2G excels in gesture generation, boasting a learning speed that is over 6$times$ faster than traditional diffusion transformers.
  arXiv  Detail & Related papers  (2024-08-06T17:29:01Z)
• DiffiT: Diffusion Vision Transformers for Image Generation [88.08529836125399]
  Vision Transformer (ViT) has demonstrated strong modeling capabilities and scalability, especially for recognition tasks. We study the effectiveness of ViTs in diffusion-based generative learning and propose a new model denoted as Diffusion Vision Transformers (DiffiT) DiffiT is surprisingly effective in generating high-fidelity images with significantly better parameter efficiency.
  arXiv  Detail & Related papers  (2023-12-04T18:57:01Z)
• Dual-Stream Attention Transformers for Sewer Defect Classification [2.5499055723658097]
  We propose a dual-stream vision transformer architecture that processes RGB and optical flow inputs for efficient sewer defect classification. Our key idea is to use self-attention regularization to harness the complementary strengths of the RGB and motion streams. By leveraging motion cues through a self-attention regularizer, we align and enhance RGB attention maps, enabling the network to concentrate on pertinent input regions.
  arXiv  Detail & Related papers  (2023-11-07T02:31:51Z)
• Dual Aggregation Transformer for Image Super-Resolution [92.41781921611646]
  We propose a novel Transformer model, Dual Aggregation Transformer, for image SR. Our DAT aggregates features across spatial and channel dimensions, in the inter-block and intra-block dual manner. Our experiments show that our DAT surpasses current methods.
  arXiv  Detail & Related papers  (2023-08-07T07:39:39Z)
• TFill: Image Completion via a Transformer-Based Architecture [69.62228639870114]
  We propose treating image completion as a directionless sequence-to-sequence prediction task. We employ a restrictive CNN with small and non-overlapping RF for token representation. In a second phase, to improve appearance consistency between visible and generated regions, a novel attention-aware layer (AAL) is introduced.
  arXiv  Detail & Related papers  (2021-04-02T01:42:01Z)

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.