Dynamic Diffusion Transformer
- URL: http://arxiv.org/abs/2410.03456v2
- Date: Wed, 9 Oct 2024 01:01:34 GMT
- Title: Dynamic Diffusion Transformer
- Authors: Wangbo Zhao, Yizeng Han, Jiasheng Tang, Kai Wang, Yibing Song, Gao Huang, Fan Wang, Yang You,
- Abstract summary: Diffusion Transformer (DiT) has demonstrated superior performance but suffers from substantial computational costs.
We propose Dynamic Diffusion Transformer (DyDiT), an architecture that dynamically adjusts its computation along both timestep and spatial dimensions during generation.
With 3% additional fine-tuning, our method reduces the FLOPs of DiT-XL by 51%, accelerates generation by 1.73, and achieves a competitive FID score of 2.07 on ImageNet.
- Score: 67.13876021157887
- License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
- Abstract: Diffusion Transformer (DiT), an emerging diffusion model for image generation, has demonstrated superior performance but suffers from substantial computational costs. Our investigations reveal that these costs stem from the static inference paradigm, which inevitably introduces redundant computation in certain diffusion timesteps and spatial regions. To address this inefficiency, we propose Dynamic Diffusion Transformer (DyDiT), an architecture that dynamically adjusts its computation along both timestep and spatial dimensions during generation. Specifically, we introduce a Timestep-wise Dynamic Width (TDW) approach that adapts model width conditioned on the generation timesteps. In addition, we design a Spatial-wise Dynamic Token (SDT) strategy to avoid redundant computation at unnecessary spatial locations. Extensive experiments on various datasets and different-sized models verify the superiority of DyDiT. Notably, with <3% additional fine-tuning iterations, our method reduces the FLOPs of DiT-XL by 51%, accelerates generation by 1.73, and achieves a competitive FID score of 2.07 on ImageNet. The code is publicly available at https://github.com/NUS-HPC-AI-Lab/ Dynamic-Diffusion-Transformer.
Related papers
- Reconstruction vs. Generation: Taming Optimization Dilemma in Latent Diffusion Models [34.15905637499148]
We propose aligning the latent space with pre-trained vision foundation models when training the visual tokenizers.
Our proposed VA-VAE significantly expands the reconstruction-generation frontier of latent diffusion models.
We build an enhanced DiT baseline with improved training strategies and architecture designs, termed LightningDiT.
arXiv Detail & Related papers (2025-01-02T18:59:40Z) - LazyDiT: Lazy Learning for the Acceleration of Diffusion Transformers [79.07412045476872]
Diffusion Transformers have emerged as the preeminent models for a wide array of generative tasks.
We show that performing the full of the model at each diffusion step is unnecessary, as some computations can be skipped by lazily reusing the results of previous steps.
We propose a lazy learning framework that efficiently leverages cached results from earlier steps to skip redundant computations.
arXiv Detail & Related papers (2024-12-17T01:12:35Z) - FlexDiT: Dynamic Token Density Control for Diffusion Transformer [31.799640242972373]
Diffusion Transformers (DiT) deliver impressive generative performance but face prohibitive computational demands.
We propose FlexDiT, a framework that dynamically adapts token density across both spatial and temporal dimensions.
Our experiments demonstrate FlexDiT's effectiveness, achieving a 55% reduction in FLOPs and a 175% improvement in inference speed.
arXiv Detail & Related papers (2024-12-08T18:59:16Z) - 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) - Alleviating Distortion in Image Generation via Multi-Resolution Diffusion Models and Time-Dependent Layer Normalization [26.926712014346432]
This paper presents innovative enhancements to diffusion models by integrating a novel multi-resolution network and time-dependent layer normalization.
Our method's efficacy is demonstrated on the class-conditional ImageNet generation benchmark, setting new state-of-the-art FID scores of 1.70 on ImageNet 256 x 256 and 2.89 on ImageNet 512 x 512.
arXiv Detail & Related papers (2024-06-13T17:59:58Z) - Learning-to-Cache: Accelerating Diffusion Transformer via Layer Caching [56.286064975443026]
We make an interesting and somehow surprising observation: the computation of a large proportion of layers in the diffusion transformer, through a caching mechanism, can be readily removed even without updating the model parameters.
We introduce a novel scheme, named Learningto-Cache (L2C), that learns to conduct caching in a dynamic manner for diffusion transformers.
Experimental results show that L2C largely outperforms samplers such as DDIM and DPM-r, alongside prior cache-based methods at the same inference speed.
arXiv Detail & Related papers (2024-06-03T18:49:57Z) - 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) - 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) - DS-Net++: Dynamic Weight Slicing for Efficient Inference in CNNs and
Transformers [105.74546828182834]
We show a hardware-efficient dynamic inference regime, named dynamic weight slicing, which adaptively slice a part of network parameters for inputs with diverse difficulty levels.
We present dynamic slimmable network (DS-Net) and dynamic slice-able network (DS-Net++) by input-dependently adjusting filter numbers of CNNs and multiple dimensions in both CNNs and transformers.
arXiv Detail & Related papers (2021-09-21T09:57:21Z)
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.