[Fix]: Fix optimizer 'ValueError' when using PReLU activation.

[KeiChiTse]

Motivation

When using parameterized activation like PReLU in FFN, current way of initializing activation functions will cause a 'ValueError' in optimizer.
i.e. ValueError: some parameters appear in more than one parameter group

Modification

Call 'build_activation_layer' function repeatedly instead of using 'self.activate' multiple times.

[CLAassistant]

All committers have signed the CLA.

[zhouzaida]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.You have signed the CLA already but the status is still pending? Let us recheck it.

Hi, thanks for your contribution. Please sign the CLA before we merge this PR.

[KeiChiTse]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.You have signed the CLA already but the status is still pending? Let us recheck it.

Hi, thanks for your contribution. Please sign the CLA before we merge this PR.

Hi! I've signed the CLA license already!

[zhouzaida]

Hi, the lint failed and you can merge the upstream 2.x to fix the error.

[KeiChiTse]

Hi, the lint failed and you can merge the upstream 2.x to fix the error.

Hi! I wonder how can I fix this mypy hook error?

[zhouzaida]

Hi, you can use the following commands to fix the mypy error.

git remote add open-mmlab git@github.com:open-mmlab/mmcv.git
git pull open-mmlab 2.x