Simple Drop-in LoRA Conditioning on Attention Layers Will Improve Your Diffusion Model

• URL: http://arxiv.org/abs/2405.03958v3
• Date: Fri, 04 Oct 2024 09:40:05 GMT
• Title: Simple Drop-in LoRA Conditioning on Attention Layers Will Improve Your Diffusion Model
• Abstract summary: Current state-of-the-art diffusion models employ U-Net architectures containing convolutional and (qkv) self-attention layers. We show that simply adding LoRA conditioning to the attention layers without changing or tuning the other parts of the U-Net architecture improves the image generation quality.
• Score: 18.138932199576917
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Current state-of-the-art diffusion models employ U-Net architectures containing convolutional and (qkv) self-attention layers. The U-Net processes images while being conditioned on the time embedding input for each sampling step and the class or caption embedding input corresponding to the desired conditional generation. Such conditioning involves scale-and-shift operations to the convolutional layers but does not directly affect the attention layers. While these standard architectural choices are certainly effective, not conditioning the attention layers feels arbitrary and potentially suboptimal. In this work, we show that simply adding LoRA conditioning to the attention layers without changing or tuning the other parts of the U-Net architecture improves the image generation quality. For example, a drop-in addition of LoRA conditioning to EDM diffusion model yields FID scores of 1.91/1.75 for unconditional and class-conditional CIFAR-10 generation, improving upon the baseline of 1.97/1.79.

Related papers

• Anchor Token Matching: Implicit Structure Locking for Training-free AR Image Editing [60.102602955261084]
  Implicit Structure Locking (ISLock) is the first training-free editing strategy for AR visual models. Our method preserves structural blueprints by dynamically aligning self-attention patterns with reference images. Our findings pioneer the way for efficient and flexible AR-based image editing, further bridging the performance gap between diffusion and autoregressive generative models.
  arXiv  Detail & Related papers  (2025-04-14T17:25:19Z)
• ARFlow: Autoregressive Flow with Hybrid Linear Attention [48.707933347079894]
  Flow models are effective at progressively generating realistic images.<n>They struggle to capture long-range dependencies during the generation process.<n>We propose integrating autoregressive modeling into flow models.
  arXiv  Detail & Related papers  (2025-01-27T14:33:27Z)
• Hierarchical Information Flow for Generalized Efficient Image Restoration [108.83750852785582]
  We propose a hierarchical information flow mechanism for image restoration, dubbed Hi-IR. Hi-IR constructs a hierarchical information tree representing the degraded image across three levels. In seven common image restoration tasks, Hi-IR achieves its effectiveness and generalizability.
  arXiv  Detail & Related papers  (2024-11-27T18:30:08Z)
• Stable Flow: Vital Layers for Training-Free Image Editing [74.52248787189302]
  Diffusion models have revolutionized the field of content synthesis and editing. Recent models have replaced the traditional UNet architecture with the Diffusion Transformer (DiT) We propose an automatic method to identify "vital layers" within DiT, crucial for image formation. Next, to enable real-image editing, we introduce an improved image inversion method for flow models.
  arXiv  Detail & Related papers  (2024-11-21T18:59:51Z)
• SeaDAG: Semi-autoregressive Diffusion for Conditional Directed Acyclic Graph Generation [83.52157311471693]
  We introduce SeaDAG, a semi-autoregressive diffusion model for conditional generation of Directed Acyclic Graphs (DAGs) Unlike conventional autoregressive generation that lacks a global graph structure view, our method maintains a complete graph structure at each diffusion step. We explicitly train the model to learn graph conditioning with a condition loss, which enhances the diffusion model's capacity to generate realistic DAGs.
  arXiv  Detail & Related papers  (2024-10-21T15:47:03Z)
• Strengthening Layer Interaction via Dynamic Layer Attention [12.341997220052486]
  Existing layer attention methods achieve layer interaction on fixed feature maps in a static manner. To restore the dynamic context representation capability of the attention mechanism, we propose a Dynamic Layer Attention architecture. Experimental results demonstrate the effectiveness of the proposed DLA architecture, outperforming other state-of-the-art methods in image recognition and object detection tasks.
  arXiv  Detail & Related papers  (2024-06-19T09:35:14Z)
• A-SDM: Accelerating Stable Diffusion through Model Assembly and Feature Inheritance Strategies [51.7643024367548]
  Stable Diffusion Model is a prevalent and effective model for text-to-image (T2I) and image-to-image (I2I) generation. This study focuses on reducing redundant computation in SDM and optimizing the model through both tuning and tuning-free methods.
  arXiv  Detail & Related papers  (2024-05-31T21:47:05Z)
• DGNet: Dynamic Gradient-Guided Network for Water-Related Optics Image Enhancement [77.0360085530701]
  Underwater image enhancement (UIE) is a challenging task due to the complex degradation caused by underwater environments. Previous methods often idealize the degradation process, and neglect the impact of medium noise and object motion on the distribution of image features. Our approach utilizes predicted images to dynamically update pseudo-labels, adding a dynamic gradient to optimize the network's gradient space.
  arXiv  Detail & Related papers  (2023-12-12T06:07:21Z)
• Semantic Image Synthesis via Class-Adaptive Cross-Attention [7.147779225315707]
  Cross-attention layers are used in place of SPADE for learning shape-style correlations and so conditioning the image generation process. Our model inherits the versatility of SPADE, at the same time obtaining state-of-the-art generation quality, as well as improved global and local style transfer.
  arXiv  Detail & Related papers  (2023-08-30T14:49:34Z)
• Hierarchical Feature Alignment Network for Unsupervised Video Object Segmentation [99.70336991366403]
  We propose a concise, practical, and efficient architecture for appearance and motion feature alignment. The proposed HFAN reaches a new state-of-the-art performance on DAVIS-16, achieving 88.7 $mathcalJ&mathcalF$ Mean, i.e., a relative improvement of 3.5% over the best published result.
  arXiv  Detail & Related papers  (2022-07-18T10:10:14Z)
• Dynamic Scene Deblurring Base on Continuous Cross-Layer Attention Transmission [6.3482616879743885]
  We introduce a new continuous cross-layer attention transmission (CCLAT) mechanism that can exploit hierarchical attention information from all the convolutional layers. Taking RDAFB as the building block, we design an effective architecture for dynamic scene deblurring named RDAFNet. Experiments on benchmark datasets show that the proposed model outperforms the state-of-the-art deblurring approaches.
  arXiv  Detail & Related papers  (2022-06-23T04:55:13Z)
• Adaptive Cross-Layer Attention for Image Restoration [10.795382525098917]
  Non-local attention processes features of each layer separately, so it risks missing correlation between features among different layers. We propose a novel Adaptive Cross-Layer Attention (ACLA) module. Two adaptive designs are proposed for ACLA: (1) adaptively selecting the keys for non-local attention at each layer; (2) automatically searching for the insertion locations for ACLA modules.
  arXiv  Detail & Related papers  (2022-03-04T22:16:18Z)
• Normalizing Flows with Multi-Scale Autoregressive Priors [131.895570212956]
  We introduce channel-wise dependencies in their latent space through multi-scale autoregressive priors (mAR) Our mAR prior for models with split coupling flow layers (mAR-SCF) can better capture dependencies in complex multimodal data. We show that mAR-SCF allows for improved image generation quality, with gains in FID and Inception scores compared to state-of-the-art flow-based models.
  arXiv  Detail & Related papers  (2020-04-08T09:07:11Z)

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.