Super Vision Transformer (Link)

This a Pytorch implementation of our paper “Super Vision Transformer”

Introduction

We present a novel training paradigm that trains only one ViT model at a time, but is capable of providing improved image recognition performance with various computational costs. Here, the trained ViT model, termed super vision transformer (SuperViT), is empowered with the versatile ability of solving incoming patches of multiple sizes as well as preserving informative tokens with multiple keeping rates (the probability of keeping tokens) to achieve a good hardware efficiency for inference.

Pre-trained Models

SuperViT(DeiT-S)

checkpoints:Google Drive
log: Google Drive

token number\keep ratio	0.5	0.7	1
8×8	73.75	75.28	75.79
10×10	77.32	78.31	78.53
12×12	78.91	79.62	79.87
14×14	79.95	80.50	80.62

SuperViT(LV-ViT-S)

checkpoints:Google Drive
log: Google Drive

token number\keep ratio	0.5	0.7	1
8×8	76.64	79.82	80.71
10×10	79.81	81.73	82.24
12×12	81.12	82.59	82.94
14×14	82.11	83.15	83.47

Requirements

python 3.9.7
pytorch 1.10.1
torchvision 0.11.2

Data Preparation

The ImageNet dataset should be prepared as follows:

ImageNet
├── train
│   ├── folder 1 (class 1)
│   ├── folder 2 (class 1)
│   ├── ...
├── val
│   ├── folder 1 (class 1)
│   ├── folder 2 (class 1)
│   ├── ...

Evaluate Pre-trained Models

Evaluate SuperViT(DeiT-S) on ImageNet

python -m torch.distributed.launch --nproc_per_node=4 main_deit.py  --model super_deit_s --batch-size 256 --data-path PATH_TO_IMAGENET --dist-eval --output PATH_TO_LOG --eval --resume PATH_TO_CHECKPOINTS

Evaluate SuperViT(LV-ViT-S) on ImageNet

python -m torch.distributed.launch --nproc_per_node=4 main_lvvit.py PATH_TO_IMAGENET --model super_lvvit_s -b 256 --native-amp --drop-path 0.1 --token-label --token-label-data PATH_TO_TOKENLABEL --token-label-size 14 --model-ema --output PATH_TO_LOG --eval --resume PATH_TO_CHECKPOINTS

Train

Train SuperViT(DeiT-S) on ImageNet

python -m torch.distributed.launch --nproc_per_node=4 main_deit.py  --model super_deit_s --batch-size 256 --data-path PATH_TO_IMAGENET --dist-eval --output PATH_TO_LOG

Train SuperViT(LV-ViT-S) on ImageNet

python -m torch.distributed.launch --nproc_per_node=4 main_lvvit.py PATH_TO_IMAGENET --model super_lvvit_s -b 256 --native-amp --drop-path 0.1 --token-label --token-label-data PATH_TO_TOKENLABEL --token-label-size 14 --model-ema --output PATH_TO_LOG