LJill
Results
2
comments of
LJill
 我在使用involution替换RCAN中的CA模块时，loss也非常大
> I replace the standard conv with involution and added bn, then the loss seems normal.But the final result is worse than edsr baseline with bn layer,even though i added...