asyml / vision-transformer-pytorch

Pytorch version of Vision Transformer (ViT) with pretrained models. This is part of CASL (https://casl-project.github.io/) and ASYML project.
https://asyml.io/
Apache License 2.0
342 stars 58 forks source link

效果不好 #1

Closed luotianhang closed 3 years ago

luotianhang commented 3 years ago

作者您好 我在cifar10 和100 上测试 效果比一般的卷积网络还要差这是为什么啊

Hhhhhhao commented 3 years ago

If you haven't fine-tune the network on CIFAR please do that. The pretrained model is trained using ImageNet. Our experiments show that even fine-tune the pretrained model with very small batch size we can still get similar accuracy as in the paper.

luotianhang commented 3 years ago

thanks for your work, appreciate it