Tokens-to-Token ViT: Training Vision Transformers from Scratch on ImageNet, arxiv
2021/03/02: update our new results. Now T2T-ViT-14 with 21.5M parameters can reach 81.5% top1-acc by training from scratch on ImageNet.
2021/02/21: our 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.
2021/02/14: update token_performer.py, now T2T-ViT-7, T2T-ViT-10, T2T-ViT-12 can be trained on 4 GPUs with 12G memory, other T2T-ViT also can be trained on 4 or 8 GPUs.
2021/01/28: init codes and upload most of the pretrained models of T2T-ViT.
Our codes are based on the official imagenet example by PyTorch and pytorch-image-models by Ross Wightman
timm, pip install timm
torch>=1.4.0
torchvision>=0.5.0
pyyaml
data prepare: ImageNet with the following folder structure:
imagenet/
....train/
........calss1/
............/ img1
........class2/
............/ img2
....val/
........calss1/
............/ img1
........class2/
............/ img2
| Model | T2T Transformer | Top1 Acc | #params | MACs | Download |
|---|---|---|---|---|---|
| T2T-ViT-14 | Performer | 81.5 | 21.5M | 5.2G | here |
| T2T-ViT-19 | Performer | 81.9 | 39.0M | 8.9G | here |
| T2T-ViT-24 | Performer | 82.2 | 64.1M | 14.1G | here |
| T2T-ViT_t-14 | Transformer | 81.7 | 21.5M | 6.1G | here |
| T2T-ViT_t-19 | Transformer | 82.4 | 39.0M | 9.8G | here |
| T2T-ViT_t-24 | Transformer | 64.1M | 15.0G | comming |
The three lite variant of T2T-ViT (Comparing with MobileNets):
| Model | T2T Transformer | Top1 Acc | #params | MACs | Download |
|---|---|---|---|---|---|
| T2T-ViT-7 | Performer | 71.7 | 4.3M | 1.2G | here |
| T2T-ViT-10 | Performer | 75.2 | 5.9M | 1.8G | here |
| T2T-ViT-12 | Performer | 76.5 | 6.9M | 2.2G | here |
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
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:
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
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
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:
If you find this repo useful, please consider citing:
@misc{yuan2021tokenstotoken,
title={Tokens-to-Token ViT: Training Vision Transformers from Scratch on ImageNet},
author={Li Yuan and Yunpeng Chen and Tao Wang and Weihao Yu and Yujun Shi and Francis EH Tay and Jiashi Feng and Shuicheng Yan},
year={2021},
eprint={2101.11986},
archivePrefix={arXiv},
primaryClass={cs.CV}
}