Inf-DiT: Upsampling Any-Resolution Image with Memory-Efficient Diffusion Transformer

TL;DR

This paper introduces Inf-DiT, a memory-efficient diffusion transformer that can generate ultra-high-resolution images up to 4096x4096 by adaptively managing memory during inference, outperforming existing models.

Contribution

The paper proposes a unidirectional block attention mechanism and an infinite super-resolution model based on DiT, enabling efficient ultra-high-resolution image generation with significantly reduced memory usage.

Findings

Achieves state-of-the-art ultra-high-resolution image generation.

Saves over 5x memory compared to UNet structures.

Successfully handles images of various shapes and resolutions.

Abstract

Diffusion models have shown remarkable performance in image generation in recent years. However, due to a quadratic increase in memory during generating ultra-high-resolution images (e.g. 4096*4096), the resolution of generated images is often limited to 1024*1024. In this work. we propose a unidirectional block attention mechanism that can adaptively adjust the memory overhead during the inference process and handle global dependencies. Building on this module, we adopt the DiT structure for upsampling and develop an infinite super-resolution model capable of upsampling images of various shapes and resolutions. Comprehensive experiments show that our model achieves SOTA performance in generating ultra-high-resolution images in both machine and human evaluation. Compared to commonly used UNet structures, our model can save more than 5x memory when generating 4096*4096 images. The project URL is https://github.com/THUDM/Inf-DiT.