You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hi, thanks for the great work. I try to use your code for my own project, however, I find it is hard to obtain the same accuracy on the previously trained task. I think the main reason is that the masked parameters do not include the batch norm layer. Do you have any idea about this behaviour?
The text was updated successfully, but these errors were encountered:
I find it is hard to obtain the same accuracy on the previously trained task too.The accuracy will reduced to half of the original size. Have you found the reason now?
Hi, thanks for the great work. I try to use your code for my own project, however, I find it is hard to obtain the same accuracy on the previously trained task. I think the main reason is that the masked parameters do not include the batch norm layer. Do you have any idea about this behaviour?
The text was updated successfully, but these errors were encountered: