Revisiting Non-Autoregressive Transformers for Efficient Image Synthesis
This work addresses the computational inefficiency of diffusion models for image synthesis, offering a more efficient alternative that is incremental in improving NATs.
The paper tackles the problem of inefficient image synthesis by revisiting non-autoregressive Transformers (NATs) to improve their performance, proposing AutoNAT, which achieves comparable results to diffusion models with significantly reduced inference cost, as validated on datasets like ImageNet-256 and MS-COCO.
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.