UniCP: A Unified Caching and Pruning Framework for Efficient Video Generation

TL;DR

UniCP introduces a unified caching and pruning framework that dynamically optimizes attention computation in diffusion transformers, significantly improving efficiency and maintaining video quality in generation tasks.

Contribution

It proposes a novel framework combining dynamic caching and pruning techniques to address computational challenges in diffusion transformer-based video generation.

Findings

Outperforms existing methods in efficiency and quality

Effectively adapts cache windows to error fluctuations

Reduces redundant attention components with PCA slicing

Abstract

Diffusion Transformers (DiT) excel in video generation but encounter significant computational challenges due to the quadratic complexity of attention. Notably, attention differences between adjacent diffusion steps follow a U-shaped pattern. Current methods leverage this property by caching attention blocks, however, they still struggle with sudden error spikes and large discrepancies. To address these issues, we propose UniCP a unified caching and pruning framework for efficient video generation. UniCP optimizes both temporal and spatial dimensions through. Error Aware Dynamic Cache Window (EDCW): Dynamically adjusts cache window sizes for different blocks at various timesteps, adapting to abrupt error changes. PCA based Slicing (PCAS) and Dynamic Weight Shift (DWS): PCAS prunes redundant attention components, and DWS integrates caching and pruning by enabling dynamic switching between pruned and cached outputs. By adjusting cache windows and pruning redundant components, UniCP enhances computational efficiency and maintains video detail fidelity. Experimental results show that UniCP outperforms existing methods in both performance and efficiency.