Awesome

ScaleNet

ScaleNet: Searching for the Model to Scale (ECCV 2022) ECVA or ArXiv

Requirements

• python >= 3.6
• 1.0.0 <= PyTorch <= 1.3.0
• torchvision >= 0.3.0
• pymoo == 0.3.0
• pymop == latest version

Super-supernet Training

• Download datasets
• Run: bash train.sh

Searching

• Run: bash search.sh

Retraining

• Please directly apply the package TIMM.

Warning

• dataloader and dataset may need to be modified for adapting your environment.

Citation

If you find this paper useful in your research, please consider citing:

@InProceedings{xie2022scalenet,
  author={Jiyang Xie and Xiu Su and Shan You and Zhanyu Ma and Fei Wang and Chen Qian},
  booktitle={European Conference on Computer Vision (ECCV)}, 
  title={{ScaleNet}: {S}earching for the Model to Scale}, 
  year={2022},
  volume={30},
}