Skip to content

Navigation Menu

Sign in
Appearance settings

Search code, repositories, users, issues, pull requests...

Provide feedback

We read every piece of feedback, and take your input very seriously.

Saved searches

Use saved searches to filter your results more quickly

Sign up
Appearance settings

[Auto] Freeze shared ZImage ControlNet transformer modules (cluster-13077-4): merged 1 of 3 PRs#8

Open
evalstate wants to merge 2 commits into
main from
merge-cluster-cluster-13077-4-20260427125508
Open

[Auto] Freeze shared ZImage ControlNet transformer modules (cluster-13077-4): merged 1 of 3 PRs #8
evalstate wants to merge 2 commits into
main from
merge-cluster-cluster-13077-4-20260427125508

Conversation

@evalstate

@evalstate evalstate commented Apr 27, 2026

Copy link
Copy Markdown
Owner

Cluster: cluster-13077-4
Base: origin/main
Branch: merge-cluster-cluster-13077-4-20260427125508

Merged:

Skipped:

  • None

Failed/conflicting:

Notes:

Next steps:

jscaldwell55 and others added 2 commits February 13, 2026 05:55
...former
The shared modules from transformer (t_embedder, all_x_embedder, cap_embedder,
noise_refiner, context_refiner, x_pad_token, cap_pad_token) are intentionally
shared references for memory efficiency. However, nothing prevented users from
accidentally training them, which would corrupt the transformer weights.
This fix explicitly freezes these shared modules with requires_grad_(False) to
enforce the intended training contract: only control_* modules (control_layers,
control_all_x_embedder, control_noise_refiner) should be trained.
Fixes huggingface#13077

Copy link
Copy Markdown
Owner Author

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

No one assigned

Labels

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

AltStyle によって変換されたページ (->オリジナル) /