Skip to content

Fix CP in-seq-split method for DeepSeek V32 and update related tests#21192

Merged
Fridge003 merged 3 commits intosgl-project:mainfrom
Fridge003:fix-v32
Mar 23, 2026
Merged

Fix CP in-seq-split method for DeepSeek V32 and update related tests#21192
Fridge003 merged 3 commits intosgl-project:mainfrom
Fridge003:fix-v32

Conversation

@Fridge003
Copy link
Copy Markdown
Collaborator

@Fridge003 Fridge003 commented Mar 23, 2026 •
edited
Loading

Motivation

  • Fix in-seq-split for CP of V32
  • Remove all the V32-Exp models in CI testcases
  • Move V32 CP tests to cp folder, and mark it as pr-test

Modifications

Accuracy Tests

Benchmarking and Profiling

Checklist

Review Process

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

@gemini-code-assist
Copy link
Copy Markdown
Contributor

gemini-code-assist Bot commented Mar 23, 2026

Summary of Changes

Hello, 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 addresses a critical bug in the context parallel (CP) implementation for DeepSeek V3.2 models by fixing an incorrect calculation of the attention context parallel size. Additionally, it streamlines the continuous integration (CI) environment by updating all test configurations to reference the official DeepSeek V3.2 model, removing dependencies on the previously used experimental version. These changes ensure more accurate model behavior and a more robust testing pipeline.

Highlights

  • Context Parallelism Fix: Corrected the calculation of attn_cp_size for DeepSeek V3.2 models to properly account for data parallelism (dp_size), resolving an in-sequence-split issue with context parallel (CP).
  • CI Test Case Standardization: Standardized the DeepSeek V3.2 model path across all relevant CI test cases, transitioning from the experimental "DeepSeek-V3.2-Exp" to the stable "DeepSeek-V3.2" variant.

🧠 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.

Footnotes

  1. 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. ↩

@Fridge003 Fridge003 changed the title Fix CP for DeepSeek V32 Fix CP in-seq-split method for DeepSeek V32 and update related tests Mar 23, 2026
gemini-code-assist[bot]
gemini-code-assist Bot reviewed Mar 23, 2026
View reviewed changes
Copy link
Copy Markdown
Contributor

@gemini-code-assist gemini-code-assist Bot left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Code Review

This pull request correctly fixes the context parallelism size calculation for DeepSeek V32 when data parallelism is active. It also standardizes the test suite by replacing the experimental DeepSeek-V3.2-Exp model with the official DeepSeek-V3.2 version across multiple test files. The logic change is sound, and the updates to the tests are appropriate. I have added one minor suggestion to improve code clarity in a test file.

Comment thread test/registered/8-gpu-models/test_deepseek_v32_cp_single_node.py Outdated
register_cuda_ci(est_time=5400, suite="nightly-8-gpu-common", nightly=True)

DEEPSEEK_V32_EXP_MODEL_PATH = "deepseek-ai/DeepSeek-V3.2-Exp"
DEEPSEEK_V32_EXP_MODEL_PATH = "deepseek-ai/DeepSeek-V3.2"
Copy link
Copy Markdown
Contributor

@gemini-code-assist gemini-code-assist Bot Mar 23, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

medium

The model path now points to the non-experimental DeepSeek-V3.2 model. For clarity and consistency with other test files, consider renaming the variable DEEPSEEK_V32_EXP_MODEL_PATH to DEEPSEEK_V32_MODEL_PATH and updating its usages in this file.

@Fridge003
Copy link
Copy Markdown
Collaborator Author

Fridge003 commented Mar 23, 2026

/rerun-stage stage-c-test-8-gpu-h200

@github-actions
Copy link
Copy Markdown
Contributor

github-actions Bot commented Mar 23, 2026

✅ Triggered stage-c-test-8-gpu-h200 to run independently (skipping dependencies).

@Fridge003
Copy link
Copy Markdown
Collaborator Author

Fridge003 commented Mar 23, 2026

/rerun-stage stage-c-test-deepep-8-gpu-h200

@github-actions
Copy link
Copy Markdown
Contributor

github-actions Bot commented Mar 23, 2026

🔗 View workflow run

@github-actions
Copy link
Copy Markdown
Contributor

github-actions Bot commented Mar 23, 2026

✅ Triggered stage-c-test-deepep-8-gpu-h200 to run independently (skipping dependencies).

@github-actions
Copy link
Copy Markdown
Contributor

github-actions Bot commented Mar 23, 2026

🔗 View workflow run

@Fridge003
Copy link
Copy Markdown
Collaborator Author

Fridge003 commented Mar 23, 2026

/rerun-stage stage-c-test-8-gpu-h200

@github-actions
Copy link
Copy Markdown
Contributor

github-actions Bot commented Mar 23, 2026

✅ Triggered stage-c-test-8-gpu-h200 to run independently (skipping dependencies).

@github-actions
Copy link
Copy Markdown
Contributor

github-actions Bot commented Mar 23, 2026

🔗 View workflow run

@Fridge003 Fridge003 merged commit ed316a2 into sgl-project:main Mar 23, 2026
64 of 72 checks passed
adityavaid pushed a commit to adityavaid/sglang that referenced this pull request Mar 24, 2026
@Fridge003 @adityavaid
Fix CP in-seq-split method for DeepSeek V32 and update related tests (s…
d767c87 
…gl-project#21192)
adityavaid pushed a commit to adityavaid/sglang that referenced this pull request Mar 24, 2026
@Fridge003 @adityavaid
Fix CP in-seq-split method for DeepSeek V32 and update related tests (s…
e67d591 
…gl-project#21192)
0-693 pushed a commit to 0-693/sglang that referenced this pull request Mar 25, 2026
@Fridge003 @0-693
Fix CP in-seq-split method for DeepSeek V32 and update related tests (s…
2796e1c 
…gl-project#21192)
JustinTong0323 pushed a commit to JustinTong0323/sglang that referenced this pull request Apr 7, 2026
@Fridge003
Fix CP in-seq-split method for DeepSeek V32 and update related tests (s…
a33aa85 
…gl-project#21192)
yhyang201 pushed a commit to yhyang201/sglang that referenced this pull request Apr 22, 2026
@Fridge003
Fix CP in-seq-split method for DeepSeek V32 and update related tests (s…
a2f919e 
…gl-project#21192)
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

1 more reviewer

@gemini-code-assist gemini-code-assist[bot] gemini-code-assist[bot] left review comments

Reviewers whose approvals may not affect merge requirements

Assignees

No one assigned

Labels

deepseek

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

1 participant

@Fridge003