Skip to content

Latest commit

 

History

History
32 lines (30 loc) · 1.05 KB

README.md

File metadata and controls

32 lines (30 loc) · 1.05 KB

EfficientNetV2-PyTorch

This is a brief PyTorch implementation of EfficientNetV2[paper], providing with experiments on ImageNet and CIFAR-10/100.
The official TensorFlow implementation is at google/automl/efficientnetv2.

Run

1. Train

sh prepare.sh
python train.py efficientnetv2-b0 cifar10

if using Distributed Data Parallel Training:

python train.py efficientnetv2-b0 cifar10 --ddp

2. Test

python eval.py efficientnetv2-b0 cifar10 --ckpt weights/best.pth.tar

Acknowledgement

With sincerely appreciation to google/automl/efficientnetv2 and lukemelas/EfficientNet-PyTorch!

Cites

@article{tan2021efficientnetv2,
  title={Efficientnetv2: Smaller models and faster training},
  author={Tan, Mingxing and Le, Quoc V},
  journal={arXiv preprint arXiv:2104.00298},
  year={2021}
}