ResearchTrend.AI
  • Papers
  • Communities
  • Events
  • Blog
  • Pricing
Papers
Communities
Social Events
Terms and Conditions
Pricing
Parameter LabParameter LabTwitterGitHubLinkedInBlueskyYoutube

© 2025 ResearchTrend.AI, All rights reserved.

  1. Home
  2. Papers
  3. 2310.00426
28
391

PixArt-ααα: Fast Training of Diffusion Transformer for Photorealistic Text-to-Image Synthesis

30 September 2023
Junsong Chen
Jincheng Yu
Chongjian Ge
Lewei Yao
Enze Xie
Yue Wu
Zhongdao Wang
James T. Kwok
Ping Luo
Huchuan Lu
Zhenguo Li
    DiffM
ArXivPDFHTML
Abstract

The most advanced text-to-image (T2I) models require significant training costs (e.g., millions of GPU hours), seriously hindering the fundamental innovation for the AIGC community while increasing CO2 emissions. This paper introduces PIXART-α\alphaα, a Transformer-based T2I diffusion model whose image generation quality is competitive with state-of-the-art image generators (e.g., Imagen, SDXL, and even Midjourney), reaching near-commercial application standards. Additionally, it supports high-resolution image synthesis up to 1024px resolution with low training cost, as shown in Figure 1 and 2. To achieve this goal, three core designs are proposed: (1) Training strategy decomposition: We devise three distinct training steps that separately optimize pixel dependency, text-image alignment, and image aesthetic quality; (2) Efficient T2I Transformer: We incorporate cross-attention modules into Diffusion Transformer (DiT) to inject text conditions and streamline the computation-intensive class-condition branch; (3) High-informative data: We emphasize the significance of concept density in text-image pairs and leverage a large Vision-Language model to auto-label dense pseudo-captions to assist text-image alignment learning. As a result, PIXART-α\alphaα's training speed markedly surpasses existing large-scale T2I models, e.g., PIXART-α\alphaα only takes 10.8% of Stable Diffusion v1.5's training time (675 vs. 6,250 A100 GPU days), saving nearly \300,000 (\26,000 vs. \320,000)andreducing90emissions.Moreover,comparedwithalargerSOTAmodel,RAPHAEL,ourtrainingcostismerely1320,000) and reducing 90% CO2 emissions. Moreover, compared with a larger SOTA model, RAPHAEL, our training cost is merely 1%. Extensive experiments demonstrate that PIXART-320,000)andreducing90emissions.Moreover,comparedwithalargerSOTAmodel,RAPHAEL,ourtrainingcostismerely1\alphaexcelsinimagequality,artistry,andsemanticcontrol.WehopePIXART− excels in image quality, artistry, and semantic control. We hope PIXART-excelsinimagequality,artistry,andsemanticcontrol.WehopePIXART−\alphawillprovidenewinsightstotheAIGCcommunityandstartupstoacceleratebuildingtheirownhigh−qualityyetlow−costgenerativemodelsfromscratch. will provide new insights to the AIGC community and startups to accelerate building their own high-quality yet low-cost generative models from scratch.willprovidenewinsightstotheAIGCcommunityandstartupstoacceleratebuildingtheirownhigh−qualityyetlow−costgenerativemodelsfromscratch.

View on arXiv
Comments on this paper