Question about layer norm in T5

[FrostML]

I notice that there is no bias and no subtraction of mean in layer norm.

I understand no bias but I'm confused about the meaning of computing variance without subtraction of mean.

Normally, we compute variance, for example:

But it's different here. Why is that?

Hoping some explanation.

[sgugger]

Please use the forums to ask such questions as we keep issues for bugs and feature requests only.

[FrostML]

Sure. have created this topic question-about-layer-norm-in-t5. I'd close this issue.