Efficient Quantization Strategies for Latent Diffusion Models

TL;DR

This paper introduces a novel quantization strategy for Latent Diffusion Models that enhances their deployability on edge devices by balancing model size reduction with performance preservation, using SQNR as a key metric.

Contribution

It proposes a combined global and local quantization approach tailored for LDMs, addressing their temporal and structural complexities for efficient PTQ.

Findings

Significant reduction in model size with minimal performance loss.

Effective quantization of sensitive and time-sensitive modules.

Improved deployment efficiency on edge devices.

Abstract

Latent Diffusion Models (LDMs) capture the dynamic evolution of latent variables over time, blending patterns and multimodality in a generative system. Despite the proficiency of LDM in various applications, such as text-to-image generation, facilitated by robust text encoders and a variational autoencoder, the critical need to deploy large generative models on edge devices compels a search for more compact yet effective alternatives. Post Training Quantization (PTQ), a method to compress the operational size of deep learning models, encounters challenges when applied to LDM due to temporal and structural complexities. This study proposes a quantization strategy that efficiently quantize LDMs, leveraging Signal-to-Quantization-Noise Ratio (SQNR) as a pivotal metric for evaluation. By treating the quantization discrepancy as relative noise and identifying sensitive part(s) of a model, we propose an efficient quantization approach encompassing both global and local strategies. The global quantization process mitigates relative quantization noise by initiating higher-precision quantization on sensitive blocks, while local treatments address specific challenges in quantization-sensitive and time-sensitive modules. The outcomes of our experiments reveal that the implementation of both global and local treatments yields a highly efficient and effective Post Training Quantization (PTQ) of LDMs.