Tokens-to-Token ViT: Training Vision Transformers from Scratch on ImageNet, ICCV 2021

Update:

2021/03/11: update our new results. Now our T2T-ViT-14 with 21.5M parameters can reach 81.5% top1-acc with 224x224 image resolution, and 83.3\% top1-acc with 384x384 resolution.

2021/02/21: T2T-ViT can be trained on most of common GPUs: 1080Ti, 2080Ti, TiTAN V, V100 stably with '--amp' (Automatic Mixed Precision). In some specifical GPU like Tesla T4, 'amp' would cause NAN loss when training T2T-ViT. If you get NAN loss in training, you can disable amp by removing '--amp' in the training scripts.

2021/01/28: release codes and upload most of the pretrained models of T2T-ViT.

Reference

If you find this repo useful, please consider citing:

@InProceedings{Yuan_2021_ICCV,
    author    = {Yuan, Li and Chen, Yunpeng and Wang, Tao and Yu, Weihao and Shi, Yujun and Jiang, Zi-Hang and Tay, Francis E.H. and Feng, Jiashi and Yan, Shuicheng},
    title     = {Tokens-to-Token ViT: Training Vision Transformers From Scratch on ImageNet},
    booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)},
    month     = {October},
    year      = {2021},
    pages     = {558-567}
}

Our codes are based on the official imagenet example by PyTorch and pytorch-image-models by Ross Wightman

1. Requirements

timm, pip install timm==0.3.4

torch>=1.4.0

torchvision>=0.5.0

pyyaml

data prepare: ImageNet with the following folder structure, you can extract imagenet by this script.

│imagenet/
├──train/
│  ├── n01440764
│  │   ├── n01440764_10026.JPEG
│  │   ├── n01440764_10027.JPEG
│  │   ├── ......
│  ├── ......
├──val/
│  ├── n01440764
│  │   ├── ILSVRC2012_val_00000293.JPEG
│  │   ├── ILSVRC2012_val_00002138.JPEG
│  │   ├── ......
│  ├── ......

2. T2T-ViT Models

The 'T2T-ViT-14, 384' means we train T2T-ViT-14 with image size of 384 x 384.

The 'T2T-ViT-24, Token Labeling' means we train T2T-ViT-24 with Token Labeling.

Usage

The way to use our pretrained T2T-ViT:

from models.t2t_vit import *
from utils import load_for_transfer_learning 

# create model
model = t2t_vit_14()

# load the pretrained weights
load_for_transfer_learning(model, /path/to/pretrained/weights, use_ema=True, strict=False, num_classes=1000)  # change num_classes based on dataset, can work for different image size as we interpolate the position embeding for different image size.

3. Validation

Test the T2T-ViT-14 (take Performer in T2T module),

Download the T2T-ViT-14, then test it by running:

CUDA_VISIBLE_DEVICES=0 python main.py path/to/data --model t2t_vit_14 -b 100 --eval_checkpoint path/to/checkpoint

The results look like:

Test: [   0/499]  Time: 2.083 (2.083)  Loss:  0.3578 (0.3578)  Acc@1: 96.0000 (96.0000)  Acc@5: 99.0000 (99.0000)
Test: [  50/499]  Time: 0.166 (0.202)  Loss:  0.5823 (0.6404)  Acc@1: 85.0000 (86.1569)  Acc@5: 99.0000 (97.5098)
...
Test: [ 499/499]  Time: 0.272 (0.172)  Loss:  1.3983 (0.8261)  Acc@1: 62.0000 (81.5000)  Acc@5: 93.0000 (95.6660)
Top-1 accuracy of the model is: 81.5%

Test the three lite variants: T2T-ViT-7, T2T-ViT-10, T2T-ViT-12 (take Performer in T2T module),

Download the T2T-ViT-7, T2T-ViT-10 or T2T-ViT-12, then test it by running:

CUDA_VISIBLE_DEVICES=0 python main.py path/to/data --model t2t_vit_7 -b 100 --eval_checkpoint path/to/checkpoint

Test the model T2T-ViT-14, 384 with 83.3\% top-1 accuracy:

CUDA_VISIBLE_DEVICES=0 python main.py path/to/data --model t2t_vit_14 --img-size 384 -b 100 --eval_checkpoint path/to/T2T-ViT-14-384 

4. Train

Train the three lite variants: T2T-ViT-7, T2T-ViT-10 and T2T-ViT-12 (take Performer in T2T module):

If only 4 GPUs are available,

CUDA_VISIBLE_DEVICES=0,1,2,3 ./distributed_train.sh 4 path/to/data --model t2t_vit_7 -b 128 --lr 1e-3 --weight-decay .03 --amp --img-size 224

The top1-acc in 4 GPUs would be slightly lower than 8 GPUs (around 0.1%-0.3% lower).

If 8 GPUs are available:

CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 ./distributed_train.sh 8 path/to/data --model t2t_vit_7 -b 64 --lr 1e-3 --weight-decay .03 --amp --img-size 224

Train the T2T-ViT-14 and T2T-ViT_t-14 (run on 4 or 8 GPUs):

CUDA_VISIBLE_DEVICES=0,1,2,3 ./distributed_train.sh 4 path/to/data --model t2t_vit_14 -b 128 --lr 1e-3 --weight-decay .05 --amp --img-size 224

CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 ./distributed_train.sh 8 path/to/data --model t2t_vit_14 -b 64 --lr 5e-4 --weight-decay .05 --amp --img-size 224

If you want to train our T2T-ViT on images with 384x384 resolution, please use '--img-size 384'.

Train the T2T-ViT-19, T2T-ViT-24 or T2T-ViT_t-19, T2T-ViT_t-24:

CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 ./distributed_train.sh 8 path/to/data --model t2t_vit_19 -b 64 --lr 5e-4 --weight-decay .065 --amp --img-size 224

5. Transfer T2T-ViT to CIFAR10/CIFAR100

We resize CIFAR10/100 to 224x224 and finetune our pretrained T2T-ViT-14/19 to CIFAR10/100 by running:

CUDA_VISIBLE_DEVICES=0,1 python transfer_learning.py --lr 0.05 --b 64 --num-classes 10 --img-size 224 --transfer-learning True --transfer-model /path/to/pretrained/T2T-ViT-19

6. Visualization

Visualize the image features of ResNet50, you can open and run the visualization_resnet.ipynb file in jupyter notebook or jupyter lab; some results are given as following:

Visualize the image features of ViT, you can open and run the visualization_vit.ipynb file in jupyter notebook or jupyter lab; some results are given as following:

Visualize attention map, you can refer to this file. A simple example by visualizing the attention map in attention block 4 and 5 is: