[Model] Fuse Z-Image's qkv_proj and gate_up_proj
#226
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Signed-off-by: Isotr0py <[email protected]>
Signed-off-by: Isotr0py <[email protected]>
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
ℹ️ About Codex in GitHub
Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
Comment on lines
+669
to
+673
| for param_name, weight_name, shard_id in stacked_params_mapping: | ||
| if weight_name not in name: | ||
| continue | ||
| name = name.replace(weight_name, param_name) | ||
| param = params_dict[name] |
There was a problem hiding this comment.
The new load_weights remaps checkpoint keys by substring, so a fused key like .to_qkv.weight or .w13.weight also matches the .to_q/.w1 patterns. The subsequent name.replace turns it into .to_qkvkv.weight/.w133.weight, which is absent from params_dict and will throw during weight loading. Any state dict saved from the fused model (the format this commit introduces) will now fail to reload. Please guard the remapping so it only applies to unfused .to_q/.w1 names.
Useful? React with 👍 / 👎.
Collaborator
|
any test result in terms of acc and speed up? |
Signed-off-by: Isotr0py <[email protected]>
Member
Author
Just benchmark manually on RTX 3090, default Main branch (45.59s)PR (45.12s)And the generated image looks reasonable: |
SamitHuang
approved these changes
Dec 8, 2025
Collaborator
|
is the runtime corresponding to omni.generate(...)? |
Member
Author
Yes, it's measured manually with these changes: +++ torch.cuda.synchronize()
+++ start_time = time.time()
images = omni.generate(
args.prompt,
height=args.height,
width=args.width,
generator=generator,
true_cfg_scale=args.cfg_scale,
num_inference_steps=args.num_inference_steps,
num_outputs_per_prompt=args.num_images_per_prompt,
)
+++ torch.cuda.synchronize()
+++ end_time = time.time()
+++ print(f"Image generation took {end_time - start_time:.2f} seconds.") |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
PLEASE FILL IN THE PR DESCRIPTION HERE ENSURING ALL CHECKLIST ITEMS (AT THE BOTTOM) HAVE BEEN CONSIDERED.
Purpose
qkv_projandgate_up_projTest Plan
Test Result
Essential Elements of an Effective PR Description Checklist
supported_models.mdandexamplesfor a new model.BEFORE SUBMITTING, PLEASE READ https://github.com/vllm-project/vllm-omni/blob/main/CONTRIBUTING.md (anything written below this line will be removed by GitHub Actions)