Autoregressive and diffusion models drive the recent breakthroughs on text-to-image generation. Despite their huge success of generating high-realistic images, a common shortcoming of these models is their high inference latency - autoregressive models run more than a thousand times successively to produce image tokens and diffusion models convert Gaussian noise into images with many hundreds of denoising steps. In this work, we explore non-autoregressive text-to-image models that efficiently generate hundreds of image tokens in parallel. We develop many model variations with different learning and inference strategies, initialized text encoders, etc. Compared with autoregressive baselines that needs to run one thousand times, our model only runs 16 times to generate images of competitive quality with an order of magnitude lower inference latency. Our non-autoregressive model with 346M parameters generates an image of 256×256 with about one second on one V100 GPU.
@article{arxiv.2312.14988,
title = {Emage: Non-Autoregressive Text-to-Image Generation},
author = {Zhangyin Feng and Runyi Hu and Liangxin Liu and Fan Zhang and Duyu Tang and Yong Dai and Xiaocheng Feng and Jiwei Li and Bing Qin and Shuming Shi},
journal= {arXiv preprint arXiv:2312.14988},
year = {2023}
}