Meissonic: Revitalizing Masked Generative Transformers for Efficient High-Resolution Text-to-Image Synthesis
Paper • 2410.08261 • Published • 52
import torch
from diffusers import DiffusionPipeline
# switch to "mps" for apple devices
pipe = DiffusionPipeline.from_pretrained("Collov-Labs/Monetico", dtype=torch.bfloat16, device_map="cuda")
prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k"
image = pipe(prompt).images[0]Monetico: An Efficient Reproduction of Meissonic for Text-to-Image Synthesis
Introduction
Similar to Meissonic, Monetico is a non-autoregressive masked image modeling text-to-image synthesis model capable of generating high-resolution images. It is designed to run efficiently on consumer-grade graphics cards.
Monetico is an efficient reproduction of Meissonic. Trained on 8 H100 GPUs for approximately one week, Monetico can generate high-quality 512x512 images that are comparable to those produced by Meissonic and SDXL.
Monetico was developed by Collov Labs. We extend our gratitude to @MeissonFlow and @viiika for their valuable advice on efficient training.
Usage
For detailed usage instructions, please refer to GitHub repository.
Citation
If you find this work helpful, please consider citing:
@article{bai2024meissonic,
title={Meissonic: Revitalizing Masked Generative Transformers for Efficient High-Resolution Text-to-Image Synthesis},
author={Bai, Jinbin and Ye, Tian and Chow, Wei and Song, Enxin and Chen, Qing-Guo and Li, Xiangtai and Dong, Zhen and Zhu, Lei and Yan, Shuicheng},
journal={arXiv preprint arXiv:2410.08261},
year={2024}
}
- Downloads last month
- 3