Qihoo-T2X: An Efficient Proxy-Tokenized Diffusion Transformer for Text-to-Any-Task

TL;DR

The paper introduces PT-DiT, a sparse attention diffusion transformer that uses proxy tokens to model global visual information efficiently, reducing computation while maintaining competitive performance in image and video generation tasks.

Contribution

It proposes a novel proxy-tokenized attention mechanism with averaging tokens for efficient global modeling in diffusion transformers, and develops the Qihoo-T2X family for various visual tasks.

Findings

Achieves up to 49% reduction in computational complexity compared to DiT.

Maintains competitive performance in image and video generation.

Introduces window and shift window attention to enhance detail modeling.

Abstract

The global self-attention mechanism in diffusion transformers involves redundant computation due to the sparse and redundant nature of visual information, and the attention map of tokens within a spatial window shows significant similarity. To address this redundancy, we propose the Proxy-Tokenized Diffusion Transformer (PT-DiT), which employs sparse representative token attention (where the number of representative tokens is much smaller than the total number of tokens) to model global visual information efficiently. Specifically, within each transformer block, we compute an averaging token from each spatial-temporal window to serve as a proxy token for that region. The global semantics are captured through the self-attention of these proxy tokens and then injected into all latent tokens via cross-attention. Simultaneously, we introduce window and shift window attention to address the limitations in detail modeling caused by the sparse attention mechanism. Building on the well-designed PT-DiT, we further develop the Qihoo-T2X family, which includes a variety of models for T2I, T2V, and T2MV tasks. Experimental results show that PT-DiT achieves competitive performance while reducing the computational complexity in both image and video generation tasks (e.g., a 49% reduction compared to DiT and a 34% reduction compared to PixArt-$\alpha$). The visual exhibition and source code of Qihoo-T2X is available at https://360cvgroup.github.io/Qihoo-T2X/.