模型描述 (Model Descriptio)

PixArt-α是一种基于Trasformer的文生图(T2I)扩散模型，其图像生成质量可与最先进的图像生成器（例如Image、SDXL甚至Midjourey）相媲美。更多详情可参照主页
模型结构图和样例结果展示如下图所示

运行环境 (Operatig eviromet)

Depedecies ad Istallatio

# Create a coda eviromet ad activate it
coda create - pixart pytho==3.9.0
coda activate pixart
pip istall torch==2.1.1 torchvisio==0.16.1 torchaudio==2.1.1 --idex-url https://dowload.pytorch.org/whl/cu118

# git cloe the origial repository
git cloe https://github.com/PixArt-alpha/PixArt-alpha.git
cd PixArt-alpha

# Istall from requiremets.txt
pip istall -r requiremets.txt

代码范例 (Code example)

参数说明：只需提供prompt，即可完成图像生成任务

from modelscope.pipelies import pipelie
iput = {'prompt': 'A small cactus with a happy face i the Sahara desert.'}

iferece = pipelie('my-pixart-task', model='aojie1997/cv_PixArt-alpha_text-to-image')
output = iferece(iput)
output.save('./result.pg')

Citatio

If you fid our work helpful for your research, please cosider citig the followig BibTeX etry.

@misc{che2023pixartalpha,
      title={PixArt-$\alpha$: Fast Traiig of Diffusio Trasformer for Photorealistic Text-to-Image Sythesis}, 
      author={Jusog Che ad Jicheg Yu ad Chogjia Ge ad Lewei Yao ad Eze Xie ad Yue Wu ad Zhogdao Wag ad James Kwok ad Pig Luo ad Huchua Lu ad Zheguo Li},
      year={2023},
      eprit={2310.00426},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}
@misc{che2024pixartdelta,
      title={PIXART-{\delta}: Fast ad Cotrollable Image Geeratio with Latet Cosistecy Models}, 
      author={Jusog Che ad Yue Wu ad Simia Luo ad Eze Xie ad Sayak Paul ad Pig Luo ad Hag Zhao ad Zheguo Li},
      year={2024},
      eprit={2401.05252},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}

模型描述 (Model Description) PixArt-α是一种基于Transformer的文生图(T2I)扩散模型，其图像生成质量可与最先进的图像生成器（例如Imagen、SDXL甚至Mid