git clone https://github.com/NVIDIA/apex
cd apex
pip install -v --disable-pip-version-check --no-cache-dir --global-option="--cpp_ext" --global-option="--cuda_ext" ./

tools/dist_train.sh configs/swin/cascade_mask_rcnn_swin_small_patch4_window7_mstrain_480-800_giou_4conv1f_adamw_3x_coco.py 8 --cfg-options model.pretrained=<PRETRAIN_MODEL> 

tools/dist_train.sh configs/swin/cascade_mask_rcnn_swin_base_patch4_window7_mstrain_480-800_giou_4conv1f_adamw_3x_coco.py 8 --cfg-options model.pretrained=<PRETRAIN_MODEL> 

@inproceedings{chang2023making,
  title={Making Vision Transformers Efficient from A Token Sparsification View},
  author={Chang, Shuning and Wang, Pichao and Lin, Ming and Wang, Fan and Zhang, David Junhao and Jin, Rong and Shou, Mike Zheng},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  pages={6195--6205},
  year={2023}
}