Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

swin训练问题 #392

Open
qiaosuobutouche opened this issue Nov 4, 2023 · 2 comments
Open

swin训练问题 #392

qiaosuobutouche opened this issue Nov 4, 2023 · 2 comments

Comments

@qiaosuobutouche
Copy link

qiaosuobutouche commented Nov 4, 2023

layumi你好,我使用swin进行训练,这是我的参数python train.py --use_swin --ft_Swin_p0.5 --erasing_p 0.5 --batchsize 16,训练20轮,我的显卡是3060laptop,6g显存,为什么loss基本不下降,并且我的val acc一直都是0.0013,不降不减,我不知道哪里出了问题,是epoch太小了么,期待您的回复

@qiaosuobutouche
Copy link
Author

qiaosuobutouche commented Nov 4, 2023

当我使用Convnext也遇到了这个问题,train Loss: 6.9570 Acc: 0.0045,val Loss: 6.5936 Acc: 0.0013,Training complete in 5m 17s,损失特别高和acc特别低,10epoch以后train loss只下降了0.0几,请问是什么原因导致的,真的很困扰我

@layumi
Copy link
Owner

layumi commented Dec 10, 2023

@qiaosuobutouche
你好 ResNet-50的话 会有问题么? 可以尝试把 lrRate改小一些。 用“--lr”
一般transformer用的lrRate都比 conv 小一些。

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants