DiverseDiT: Towards Diverse Representation Learning in Diffusion Transformers

TL;DR

This paper investigates the internal representation dynamics of Diffusion Transformers and introduces DiverseDiT, a framework that promotes diverse internal representations to improve performance and convergence in visual synthesis tasks.

Contribution

The paper reveals the importance of representation diversity in DiTs and proposes DiverseDiT, which explicitly encourages diverse features through residual connections and a diversity loss.

Findings

DiverseDiT improves performance across different backbones and sizes.

DiverseDiT accelerates convergence in diffusion models.

DiverseDiT complements existing representation learning methods.

Abstract

Recent breakthroughs in Diffusion Transformers (DiTs) have revolutionized the field of visual synthesis due to their superior scalability. To facilitate DiTs' capability of capturing meaningful internal representations, recent works such as REPA incorporate external pretrained encoders for representation alignment. However, the underlying mechanisms governing representation learning within DiTs are not well understood. To this end, we first systematically investigate the representation dynamics of DiTs. Through analyzing the evolution and influence of internal representations under various settings, we reveal that representation diversity across blocks is a crucial factor for effective learning. Based on this key insight, we propose DiverseDiT, a novel framework that explicitly promotes representation diversity. DiverseDiT incorporates long residual connections to diversify input representations across blocks and a representation diversity loss to encourage blocks to learn distinct features. Extensive experiments on ImageNet 256x256 and 512x512 demonstrate that our DiverseDiT yields consistent performance gains and convergence acceleration when applied to different backbones with various sizes, even when tested on the challenging one-step generation setting. Furthermore, we show that DiverseDiT is complementary to existing representation learning techniques, leading to further performance gains. Our work provides valuable insights into the representation learning dynamics of DiTs and offers a practical approach for enhancing their performance.