LiteVAE: Lightweight and Efficient Variational Autoencoders for Latent Diffusion Models
This work addresses the scalability and efficiency problem for researchers and practitioners using latent diffusion models in image generation, though it is incremental as it builds on existing autoencoder designs.
The paper tackles the computational inefficiency of autoencoders in latent diffusion models by introducing LiteVAE, which uses a 2D discrete wavelet transform to reduce encoder parameters by six-fold without sacrificing output quality, and a larger model outperforms existing VAEs on metrics like rFID and LPIPS.
Advances in latent diffusion models (LDMs) have revolutionized high-resolution image generation, but the design space of the autoencoder that is central to these systems remains underexplored. In this paper, we introduce LiteVAE, a new autoencoder design for LDMs, which leverages the 2D discrete wavelet transform to enhance scalability and computational efficiency over standard variational autoencoders (VAEs) with no sacrifice in output quality. We investigate the training methodologies and the decoder architecture of LiteVAE and propose several enhancements that improve the training dynamics and reconstruction quality. Our base LiteVAE model matches the quality of the established VAEs in current LDMs with a six-fold reduction in encoder parameters, leading to faster training and lower GPU memory requirements, while our larger model outperforms VAEs of comparable complexity across all evaluated metrics (rFID, LPIPS, PSNR, and SSIM).