Cross-Attention Makes Inference Cumbersome in Text-to-Image Diffusion Models

Wentian Zhang, L, Jinheng Xie, Francesco Faccio, Mike Zheng Shou, Jürgen Schmidhuber

Cross-Attention Makes Inference Cumbersome in Text-to-Image Diffusion Models: 12 upvotes on Hugging Face Daily Papers, #7 of 8 papers on 2024-04-04. Day-by-day upvote history.

This study explores the role of cross-attention during inference in text-conditional diffusion models. We find that cross-attention outputs converge to a fixed point after few inference steps. Accordingly, the time point of convergence naturally divides the entire inference process into two stages: an initial semantics-planning stage, during which, the model relies on cross-attention to plan text-oriented visual semantics, and a subsequent fidelity-improving stage, during which the model tries to generate images from previously planned semantics. Surprisingly, ignoring text conditions in the fidelity-improving stage not only reduces computation complexity, but also maintains model performance. This yields a simple and training-free method called TGATE for efficient generation, which caches the cross-attention output once it converges and keeps it fixed during the remaining inference steps. Our empirical study on the MS-COCO validation set confirms its effectiveness. The source code of TGATE is available at https://github.com/HaozheLiu-ST/T-GATE.

Paper page on Hugging Face · arXiv

Data: hysts-bot-data/daily-papers-stats and the Daily Papers API. Open data: tardellirs/paper-pulse-data. Sister project: Model Pulse, the download history of every model on the Hub.