Collaborative Decoding Makes Visual Auto-Regressive Modeling Efficient

TL;DR

This paper introduces Collaborative Decoding (CoDe), a strategy that significantly improves the efficiency of Visual Auto-Regressive models by splitting inference between a large and small model, reducing memory and computation with minimal quality loss.

Contribution

The paper proposes a novel collaborative decoding method for VAR models that enhances efficiency by leveraging scale-specific generation roles between two models.

Findings

Achieves 1.7x speedup and 50% memory reduction with minimal quality loss.

Further decreases in drafting steps lead to 2.9x acceleration at 41 images/sec.

Maintains high image quality with negligible FID increase.

Abstract

In the rapidly advancing field of image generation, Visual Auto-Regressive (VAR) modeling has garnered considerable attention for its innovative next-scale prediction approach. This paradigm offers substantial improvements in efficiency, scalability, and zero-shot generalization. Yet, the inherently coarse-to-fine nature of VAR introduces a prolonged token sequence, leading to prohibitive memory consumption and computational redundancies. To address these bottlenecks, we propose Collaborative Decoding (CoDe), a novel efficient decoding strategy tailored for the VAR framework. CoDe capitalizes on two critical observations: the substantially reduced parameter demands at larger scales and the exclusive generation patterns across different scales. Based on these insights, we partition the multi-scale inference process into a seamless collaboration between a large model and a small model. The large model serves as the 'drafter', specializing in generating low-frequency content at smaller scales, while the smaller model serves as the 'refiner', solely focusing on predicting high-frequency details at larger scales. This collaboration yields remarkable efficiency with minimal impact on quality: CoDe achieves a 1.7x speedup, slashes memory usage by around 50%, and preserves image quality with only a negligible FID increase from 1.95 to 1.98. When drafting steps are further decreased, CoDe can achieve an impressive 2.9x acceleration ratio, reaching 41 images/s at 256x256 resolution on a single NVIDIA 4090 GPU, while preserving a commendable FID of 2.27. The code is available at https://github.com/czg1225/CoDe