Vision Transformer with Deformable Attention (Jittor)
This repository contains a simple implementation of Vision Transformer with Deformable Attention [arXiv].
Currently, we only release the code of models and the training scripts are under development including advance data augmentations and mixed precision training.
Pytorch version: Github
Dependencies
- NVIDIA GPU + CUDA 11.1 + cuDNN 8.0.3
- Python 3.7 (Recommend to use Anaconda)
- jittor == 1.3.1.40
- jimm
TODO
- Training scripts with advance data augmentations.
Citation
If you find our work is useful in your research, please consider citing:
@misc{xia2022vision,
title={Vision Transformer with Deformable Attention},
author={Zhuofan Xia and Xuran Pan and Shiji Song and Li Erran Li and Gao Huang},
year={2022},
eprint={2201.00520},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
