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

关于BN层冻结与batch_size过小的问题 #13

Open
jintaiyang777 opened this issue Apr 6, 2019 · 2 comments
Open

关于BN层冻结与batch_size过小的问题 #13

jintaiyang777 opened this issue Apr 6, 2019 · 2 comments

Comments

@jintaiyang777
Copy link

请教一个问题,训练集图片这么大,训练时的batch_size应该是比较小的,那么BN层需要冻结吗,当时您是怎么做的

@zlckanata
Copy link
Owner

如果单卡batchsize只能设置为1,那么一定需要对BN进行特殊操作,例如跨卡BN、冻结BN等。
如果单卡batchsize不小于2,那么BN还是有意义的,可以选择正常使用。

@jintaiyang777
Copy link
Author

好的,多谢~,我今天下午也刚实验了下,在别的小数据集上训练的,我的单卡batchsize为3,将BN冻结后最后效果有一点点提升。

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