[wip] sync with upstream zImage by yhyang201 · Pull Request #17822 · sgl-project/sglang

yhyang201 · 2026-01-27T11:45:57Z

Motivation

Modifications

Accuracy Tests

Benchmarking and Profiling

Checklist

Format your code according to the Format code with pre-commit.
Add unit tests according to the Run and add unit tests.
Update documentation according to Write documentations.
Provide accuracy and speed benchmark results according to Test the accuracy and Benchmark the speed.
Follow the SGLang code style guidance.

Review Process

Ping Merge Oncalls to start the PR flow. See the PR Merge Process.
Get approvals from CODEOWNERS and other reviewers.
Trigger CI tests with comments or contact authorized users to do so.
- /tag-run-ci-label, /rerun-failed-ci, /tag-and-rerun-ci
After green CI and required approvals, ask Merge Oncalls to merge.

gemini-code-assist · 2026-01-27T11:46:16Z

Summary of Changes

Hello @yhyang201, I'm Gemini Code Assist¹! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed!

This pull request synchronizes the Z-Image model integration with its upstream source by introducing distinct configurations for 'Turbo' and standard Z-Image models. It enhances the Z-Image pipeline with negative conditioning capabilities and refactors sampling parameters to provide separate defaults for each model variant, ensuring accurate model detection and configuration within the system.

Highlights

Negative Conditioning Support: Added a new method prepare_neg_cond_kwargs to the ZImagePipelineConfig to support negative conditioning, enhancing control over image generation.
Z-Image Sampling Parameter Refactoring: Refactored Z-Image sampling parameters by renaming the original ZImageSamplingParams to ZImageTurboSamplingParams and introducing a new ZImageSamplingParams class with distinct default values, differentiating between 'Turbo' and standard Z-Image models.
Updated Model Registration: Modified the model registry to correctly associate the new ZImageTurboSamplingParams with 'Z-Image-Turbo' models and the new ZImageSamplingParams with standard 'Z-Image' models, including refined model detection logic.

🧠 New Feature in Public Preview: You can now enable Memory to help Gemini Code Assist learn from your team's feedback. This makes future code reviews more consistent and personalized to your project's style. Click here to enable Memory in your admin console.

Using Gemini Code Assist

The full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips.

Invoking Gemini

You can request assistance from Gemini at any point by creating a comment using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands on the current page.

Feature	Command	Description
Code Review	`/gemini review`	Performs a code review for the current pull request in its current state.
Pull Request Summary	`/gemini summary`	Provides a summary of the current pull request in its current state.
Comment	@gemini-code-assist	Responds in comments when explicitly tagged, both in pull request comments and review comments.
Help	`/gemini help`	Displays a list of available commands.

Customization

To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a .gemini/ folder in the base of the repository. Detailed instructions can be found here.

Limitations & Feedback

Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for Github and other Google products, sign up here.

You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension.

Review the Privacy Notices, Generative AI Prohibited Use Policy, Terms of Service, and learn how to configure Gemini Code Assist in GitHub here. Gemini can make mistakes, so double check it and use code with caution. ↩

gemini-code-assist

Code Review

This pull request introduces a new prepare_neg_cond_kwargs method in zimage.py and refactors sampling parameters in sample/zimage.py to distinguish between 'Turbo' and standard ZImage models. The changes in registry.py correctly reflect these new configurations. However, there is a critical issue in the prepare_neg_cond_kwargs method where positive prompt embeddings are used instead of negative ones, which will lead to incorrect model behavior. Additionally, the default negative prompt for the new ZImageSamplingParams could be clarified.

gemini-code-assist · 2026-01-27T11:49:08Z

python/sglang/multimodal_gen/configs/pipeline_configs/zimage.py

+    def prepare_neg_cond_kwargs(self, batch, device, rotary_emb, dtype):
+        return {
+            "freqs_cis": self.get_freqs_cis(
+                batch.prompt_embeds[0],


The prepare_neg_cond_kwargs method is intended for negative conditioning, but it currently uses batch.prompt_embeds[0] to generate freqs_cis. This should be batch.negative_prompt_embeds[0] to correctly apply negative conditioning. Using positive prompt embeddings here will lead to incorrect model behavior and potentially degraded generation quality.

Suggested change

batch.prompt_embeds[0],

batch.negative_prompt_embeds[0],

gemini-code-assist · 2026-01-27T11:49:08Z

python/sglang/multimodal_gen/configs/sample/zimage.py

+    num_inference_steps: int = 50
+
+    num_frames: int = 1
+    negative_prompt: str = " "


The negative_prompt for ZImageSamplingParams is set to a single space " ". This deviates significantly from the detailed default negative prompt in the base SamplingParams class. If the intention is to effectively disable negative prompting or use a minimal one, consider using None or an empty string "" for clarity, or add a comment explaining why a single space is specifically required for this model.

gemini-code-assist · 2026-01-28T04:07:48Z

Warning

You have reached your daily quota limit. Please wait up to 24 hours and I will start processing your requests again!

yhyang201 · 2026-01-28T04:07:52Z

/tag-and-rerun-ci

mickqian · 2026-01-28T16:25:10Z

/rerun-failed-ci

mickqian · 2026-01-29T01:09:47Z

python/sglang/multimodal_gen/registry.py

+        hf_model_paths=[
+            "Tongyi-MAI/Z-Image",
+        ],
+        model_detectors=[


could it works without these two model_detectors lines?

- fix CI formatting issues - bump torchada version - revert incorrect changes Signed-off-by: yafeng.li <yafeng.li@mthreads.com> [diffusion] model: sync with upstream z-Image (sgl-project#17822)

upd

a141abf

github-actions bot added the diffusion SGLang Diffusion label Jan 27, 2026

gemini-code-assist bot reviewed Jan 27, 2026

View reviewed changes

upd

b4b9691

yhyang201 marked this pull request as ready for review January 28, 2026 04:07

yhyang201 requested a review from mickqian as a code owner January 28, 2026 04:07

github-actions bot added the run-ci label Jan 28, 2026

Merge branch 'main' into zimage

fd63638

mickqian added the high priority label Jan 28, 2026

Merge branch 'main' into zimage

f45e8b1

mickqian reviewed Jan 29, 2026

View reviewed changes

mickqian approved these changes Jan 29, 2026

View reviewed changes

mickqian merged commit 3c2f4c7 into sgl-project:main Jan 29, 2026
176 of 188 checks passed

charlesHsuGG pushed a commit to charlesHsuGG/sglang that referenced this pull request Jan 30, 2026

[diffusion] model: sync with upstream z-Image (sgl-project#17822)

6dc1195

Chen-0210 pushed a commit to Chen-0210/sglang that referenced this pull request Jan 30, 2026

[diffusion] model: sync with upstream z-Image (sgl-project#17822)

4cbd82d

sfiisf pushed a commit to sfiisf/sglang that referenced this pull request Feb 5, 2026

[diffusion] model: sync with upstream z-Image (sgl-project#17822)

26d0096

Johnsonms pushed a commit to Johnsonms/sglang that referenced this pull request Feb 14, 2026

[diffusion] model: sync with upstream z-Image (sgl-project#17822)

ae9497c

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[wip] sync with upstream zImage #17822

[wip] sync with upstream zImage #17822
mickqian merged 4 commits intosgl-project:mainfrom
yhyang201:zimage

yhyang201 commented Jan 27, 2026

Uh oh!

gemini-code-assist bot commented Jan 27, 2026

Uh oh!

gemini-code-assist bot left a comment

Uh oh!

gemini-code-assist bot Jan 27, 2026

Uh oh!

gemini-code-assist bot Jan 27, 2026

Uh oh!

gemini-code-assist bot commented Jan 28, 2026

Uh oh!

yhyang201 commented Jan 28, 2026

Uh oh!

mickqian commented Jan 28, 2026

Uh oh!

mickqian Jan 29, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

yhyang201 commented Jan 27, 2026

Motivation

Modifications

Accuracy Tests

Benchmarking and Profiling

Checklist

Review Process

Uh oh!

gemini-code-assist bot commented Jan 27, 2026

Summary of Changes

Highlights

Footnotes

Uh oh!

gemini-code-assist bot left a comment

Choose a reason for hiding this comment

Code Review

Uh oh!

gemini-code-assist bot Jan 27, 2026

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist bot Jan 27, 2026

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist bot commented Jan 28, 2026

Uh oh!

yhyang201 commented Jan 28, 2026

Uh oh!

mickqian commented Jan 28, 2026

Uh oh!

mickqian Jan 29, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants