Applying the gamma parameter of RMSNorm

[Gyu007]

Hello.

Thank you for releasing the code of SD3.5.

I was analyzing the code to apply SD3.5 to my research.
While analyzing the code, I saw that RMSNorm defined in the DismantledBlock class in mmditx.py specifies
the elementwise_affine parameter as false by default.
This means that the scale parameter of the normalization layer is not applied.
Can you tell me why it is set to false? And did you actually train it with false as the training method?

Thank you!