Home

Awesome

Continuous Speculative Decoding for Autoregressive Image Generation <br><sub>Official PyTorch Implementation</sub>

arXiv 

<p align="center"> <img src="demo/visual.png" width="720"> </p>

This is a PyTorch/GPU implementation of the paper Continuous Speculative Decoding for Autoregressive Image Generation:

@article{wang2024continuous,
  title={Continuous Speculative Decoding for Autoregressive Image Generation},
  author={Wang, Zili and Zhang, Robert and Ding, Kun and Yang, Qi and Li, Fei and Xiang, Shiming},
  journal={arXiv preprint arXiv:2411.11925},
  year={2024}
}

Preparation

Please refer to the MAR repo for basic installation of dependencies and pretrained checkpoints.

Usage

Demo

Run our interactive visualization demo using Colab notebook!

Acknowledgements

Codes in this repo is largely based on MAR. We thank the authors of MAR for the contributions to the community.

Contact

If you have any questions, feel free to contact me through email (wangzili2022@ia.ac.cn). Enjoy!