Revisiting Non-Autoregressive Transformers for Efficient Image Synthesis

TL;DR

This paper re-evaluates non-autoregressive Transformers for image synthesis, proposing AutoNAT to optimize training and inference strategies, achieving diffusion-model-level performance with lower computational costs.

Contribution

We introduce AutoNAT, an automatic framework that optimizes NAT strategies, significantly improving their performance to match diffusion models in image synthesis.

Findings

AutoNAT outperforms previous NATs on benchmark datasets.

AutoNAT achieves comparable results to diffusion models with less inference cost.

The method is validated on four major datasets, demonstrating broad effectiveness.

Abstract

The field of image synthesis is currently flourishing due to the advancements in diffusion models. While diffusion models have been successful, their computational intensity has prompted the pursuit of more efficient alternatives. As a representative work, non-autoregressive Transformers (NATs) have been recognized for their rapid generation. However, a major drawback of these models is their inferior performance compared to diffusion models. In this paper, we aim to re-evaluate the full potential of NATs by revisiting the design of their training and inference strategies. Specifically, we identify the complexities in properly configuring these strategies and indicate the possible sub-optimality in existing heuristic-driven designs. Recognizing this, we propose to go beyond existing methods by directly solving the optimal strategies in an automatic framework. The resulting method, named AutoNAT, advances the performance boundaries of NATs notably, and is able to perform comparably with the latest diffusion models at a significantly reduced inference cost. The effectiveness of AutoNAT is validated on four benchmark datasets, i.e., ImageNet-256 & 512, MS-COCO, and CC3M. Our code is available at https://github.com/LeapLabTHU/ImprovedNAT.