remove silu for CogView4

[lambertwjh]

What does this PR do?

Fixes # (issue)

Before submitting

• This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case).
• Did you read the contributor guideline?
• Did you read our philosophy doc (important for complex PRs)?
• Was this discussed/approved via a GitHub issue or the forum? Please add a link to it if that's the case.
• Did you make sure to update the documentation with your changes? Here are the
  documentation guidelines, and
  here are tips on formatting docstrings.
• Did you write any new necessary tests?

Who can review?

Anyone in the community is free to review the PR once the tests have passed. Feel free to tag
members/contributors who may be interested in your PR.

[lambertwjh]

This is a small CogView4 alignment fix：
What this PR does

CogView4: remove the SiLU right before the final AdaLN.
Add a small switch use_silu (default True) to AdaLayerNormContinuous.
In CogView4Transformer2DModel, set norm_out = AdaLayerNormContinuous(..., use_silu=False) and keep SiLU only for the block path (temb_blocks = F.silu(temb_raw)) while feeding the raw temb_raw to the final AdaLN.

Why

In Megatron (our reference), adaln_final is a plain Linear (no activation).
In diffusers, temb was passed through F.silu before norm_out, which makes the final modulation differ from Megatron and causes hidden-state @@mismatches.
This PR aligns diffusers’ CogView4 with Megatron’s behavior.

[zRzRzRzRzRzRzR]

@a-r-r-o-w can you check this

[a-r-r-o-w]

Instead of making this change here, can we just copy the layer into the CogView transformer files where required and use the correct code?

[lambertwjh]

“CogView4-only fix: final AdaLN without SiLU to match Megatron; generic normalization remains unchanged.”
please check again @a-r-r-o-w

[a-r-r-o-w]

Thanks @lambertwjh, looks good. I'm adding some changes to update the layer names.

[a-r-r-o-w]

@bot style

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

[a-r-r-o-w]

@bot /style

[github-actions]

Style bot fixed some files and pushed the changes.

[a-r-r-o-w]

@lambertwjh @zRzRzRzRzRzRzR Just to confirm, do the changes for the layer name look good to you? If so, I'll merge this :)

[lambertwjh]

Looks good to me. The rename to CogView4AdaLayerNormContinuous is clearer and matches the CogView4-only usage. Thanks! @a-r-r-o-w