SNED: Superposition Network Architecture Search for Efficient Video Diffusion Model

TL;DR

SNED introduces a superposition network architecture search method that optimizes video diffusion models for efficiency, enabling high-quality video synthesis at various resolutions and model sizes with reduced computational costs.

Contribution

The paper proposes a novel supernet training paradigm with sampling warm-up for efficient architecture search in video diffusion models, applicable to both pixel and latent spaces.

Findings

Achieves consistent video generation across multiple resolutions.

Supports a wide range of model sizes from 640M to 1.6B parameters.

Demonstrates high efficiency with comparable results to larger models.

Abstract

While AI-generated content has garnered significant attention, achieving photo-realistic video synthesis remains a formidable challenge. Despite the promising advances in diffusion models for video generation quality, the complex model architecture and substantial computational demands for both training and inference create a significant gap between these models and real-world applications. This paper presents SNED, a superposition network architecture search method for efficient video diffusion model. Our method employs a supernet training paradigm that targets various model cost and resolution options using a weight-sharing method. Moreover, we propose the supernet training sampling warm-up for fast training optimization. To showcase the flexibility of our method, we conduct experiments involving both pixel-space and latent-space video diffusion models. The results demonstrate that our framework consistently produces comparable results across different model options with high efficiency. According to the experiment for the pixel-space video diffusion model, we can achieve consistent video generation results simultaneously across 64 x 64 to 256 x 256 resolutions with a large range of model sizes from 640M to 1.6B number of parameters for pixel-space video diffusion models.