PaddleMAE

Masked Autoencoders Are Scalable Vision Learners

Dataset: ImageNet1k Target: ViT-B，imagenet1k val 83.6%

V100 CUDA 10.2

conda create -n paddle-mae python=3.7
conda activate paddle-mae
pip install -r ./requirements.txt