[Bugfix] fix the wrong use of extract_slice and insert_slice by wangx700 · Pull Request #6956 · vllm-project/vllm-ascend

wangx700 · 2026-03-03T08:40:03Z

What this PR does / why we need it?

fix the wrong use of extract_slice and insert_slice. In the TritonAscend main branch, extract_slice and insert_slice are located in the triton.language.extra.cann.extension library, rather than in the triton.language library

Does this PR introduce any user-facing change?

How was this patch tested?

vLLM version: v0.16.0
vLLM main: vllm-project/vllm@15d76f7

…end main branch, extract_slice and insert_slice are located in the triton.language.extra.cann.extension library, rather than in the triton.language library Signed-off-by: wangx700 <wangxin700@huawei.com>

gemini-code-assist · 2026-03-03T08:40:19Z

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 resolves an issue where Triton's extract_slice and insert_slice functions were incorrectly referenced in vllm_ascend modules, specifically for the TritonAscend environment. By introducing conditional imports, the change ensures that the correct versions of these functions are used, enhancing the adaptability and stability of the codebase across different Triton configurations without requiring manual adjustments.

Highlights

Conditional Import for Triton Slicing Functions: Implemented a conditional import mechanism for extract_slice and insert_slice functions. The code now attempts to import these functions from triton.language.extra.cann.extension first, falling back to triton.language if the specialized extension is not available. This ensures compatibility with the TritonAscend main branch where these functions are located in a different module.
Updated Usage of Slicing Functions: All direct calls to tl.extract_slice and tl.insert_slice across the affected files have been replaced with the newly defined extract_slice and insert_slice variables, which dynamically resolve to the correct implementation based on the conditional import.

🧠 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.

Changelog

vllm_ascend/ops/triton/activation/swiglu_quant.py
- Added conditional import logic for extract_slice to support triton.language.extra.cann.extension.
- Updated all usages of tl.extract_slice to use the newly defined extract_slice variable.
vllm_ascend/ops/triton/fla/solve_tril.py
- Added conditional import logic for insert_slice and extract_slice to support triton.language.extra.cann.extension.
- Updated all usages of tl.insert_slice and tl.extract_slice to use the newly defined insert_slice and extract_slice variables.
vllm_ascend/ops/triton/linearnorm/split_qkv_rmsnorm_rope.py
- Added conditional import logic for insert_slice and extract_slice to support triton.language.extra.cann.extension.
- Updated all usages of tl.insert_slice and tl.extract_slice to use the newly defined insert_slice and extract_slice variables.

Activity

No specific activity has been recorded for this pull request yet.

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. ↩

github-actions · 2026-03-03T08:40:38Z

👋 Hi! Thank you for contributing to the vLLM Ascend project. The following points will speed up your PR merge:‌‌

A PR should do only one thing, smaller PRs enable faster reviews.
Every PR should include unit tests and end-to-end tests ‌to ensure it works and is not broken by other future PRs.
Write the commit message by fulfilling the PR description to help reviewer and future developers understand.

If CI fails, you can run linting and testing checks locally according Contributing and Testing.

gemini-code-assist

Code Review

The pull request correctly identifies and addresses the issue of extract_slice and insert_slice being located in a specific Triton extension library (triton.language.extra.cann.extension) rather than the standard triton.language module in the TritonAscend main branch. The changes introduce conditional imports to use the extension's functions if available, falling back to the standard tl module otherwise. This improves compatibility across different Triton environments.

However, the fallback mechanism in the except ImportError block assumes that tl.extract_slice and tl.insert_slice will always exist in the triton.language module. Given the PR description stating these functions are "rather than in the triton.language library" for the specific branch, this assumption might be incorrect and could lead to AttributeError if the standard tl module does not define these functions. A more robust fallback is suggested to explicitly check for the existence of these functions in tl or raise a more informative error.

gemini-code-assist · 2026-03-03T08:41:54Z

+try:
+    import triton.language.extra.cann.extension as extension
+    extract_slice = extension.extract_slice
+except ImportError:
+    extract_slice = tl.extract_slice


The current try-except ImportError block assumes that tl.extract_slice will always be available if triton.language.extra.cann.extension cannot be imported. However, the PR description indicates that these functions might not be in the standard triton.language library for the TritonAscend main branch. If tl.extract_slice is indeed missing, this fallback will result in an AttributeError at runtime.

To make this more robust, consider explicitly checking for the existence of tl.extract_slice or handling AttributeError in the fallback, or raising a more specific error if neither source provides the function.

try: import triton.language.extra.cann.extension as extension extract_slice = extension.extract_slice except ImportError: if hasattr(tl, 'extract_slice'): extract_slice = tl.extract_slice else: raise RuntimeError("Neither triton.language.extra.cann.extension.extract_slice nor tl.extract_slice is available.")

gemini-code-assist · 2026-03-03T08:41:55Z

+try:
+    import triton.language.extra.cann.extension as extension
+    insert_slice = extension.insert_slice
+    extract_slice = extension.extract_slice
+except ImportError:
+    insert_slice = tl.insert_slice
+    extract_slice = tl.extract_slice



Similar to the issue in swiglu_quant.py, the fallback in the except ImportError block assumes tl.insert_slice and tl.extract_slice are always present. If these functions are not part of the standard triton.language module in some environments, this could lead to an AttributeError.

It's recommended to add a check for the existence of these functions in tl or to catch AttributeError to provide a more graceful failure or a clearer error message.

Suggested change

try:

import triton.language.extra.cann.extension as extension

insert_slice = extension.insert_slice

extract_slice = extension.extract_slice

except ImportError:

insert_slice = tl.insert_slice

extract_slice = tl.extract_slice

try:

import triton.language.extra.cann.extension as extension

insert_slice = extension.insert_slice

extract_slice = extension.extract_slice

except ImportError:

if hasattr(tl, 'insert_slice') and hasattr(tl, 'extract_slice'):

insert_slice = tl.insert_slice

extract_slice = tl.extract_slice

else:

raise RuntimeError("Neither triton.language.extra.cann.extension nor tl provides insert_slice/extract_slice.")

gemini-code-assist · 2026-03-03T08:41:55Z

+try:
+    import triton.language.extra.cann.extension as extension
+    insert_slice = extension.insert_slice
+    extract_slice = extension.extract_slice
+except ImportError:
+    insert_slice = tl.insert_slice
+    extract_slice = tl.extract_slice



The conditional import for insert_slice and extract_slice has the same potential issue as identified in the other files. If triton.language.extra.cann.extension fails to import, the fallback to tl.insert_slice and tl.extract_slice might fail with an AttributeError if these functions are not defined in the standard triton.language module for the current environment.

Please implement a more robust fallback mechanism to ensure these functions are available or to provide a clear error if they are not.

try: import triton.language.extra.cann.extension as extension insert_slice = extension.insert_slice extract_slice = extension.extract_slice except ImportError: if hasattr(tl, 'insert_slice') and hasattr(tl, 'extract_slice'): insert_slice = tl.insert_slice extract_slice = tl.extract_slice else: raise RuntimeError("Neither triton.language.extra.cann.extension nor tl provides insert_slice/extract_slice.")

Signed-off-by: wangx700 <wangxin700@huawei.com>

linfeng-yuan

#6937 has already been merged to avoid these cases and keep compatibility between triton-ascend 3.2 and newest one. Please check that whether it can help with your case~

Ronald1995 · 2026-03-04T02:14:38Z

+
+    insert_slice = extension.insert_slice
+    extract_slice = extension.extract_slice
+except ImportError:


you could move the same logic to triton_utils.py

fix the wrong use of extract_slice and insert_slice. In the TritonAsc…

b531798

…end main branch, extract_slice and insert_slice are located in the triton.language.extra.cann.extension library, rather than in the triton.language library Signed-off-by: wangx700 <wangxin700@huawei.com>

wangx700 requested review from realliujiaxu, whx-sjtu and zzzzwwjj as code owners March 3, 2026 08:40

github-actions bot added the module:ops label Mar 3, 2026

gemini-code-assist bot reviewed Mar 3, 2026

View reviewed changes

wangx700 added 2 commits March 3, 2026 17:28

fix ruff format.

560ec28

Signed-off-by: wangx700 <wangxin700@huawei.com>

fix mypy test with "# type: ignore"

0feb4ea

Signed-off-by: wangx700 <wangxin700@huawei.com>

wangx700 changed the title ~~fix the wrong use of extract_slice and insert_slice~~ [Bugfix]fix the wrong use of extract_slice and insert_slice Mar 3, 2026

wangx700 changed the title ~~[Bugfix]fix the wrong use of extract_slice and insert_slice~~ [Bugfix] fix the wrong use of extract_slice and insert_slice Mar 3, 2026

linfeng-yuan reviewed Mar 3, 2026

View reviewed changes

gcanlin mentioned this pull request Mar 3, 2026

[Release]: Release checklist for v0.16.0rc1 #6970

Closed

33 tasks

Ronald1995 reviewed Mar 4, 2026

View reviewed changes

wangx700 closed this Mar 4, 2026

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[Bugfix] fix the wrong use of extract_slice and insert_slice#6956

[Bugfix] fix the wrong use of extract_slice and insert_slice#6956
wangx700 wants to merge 3 commits intovllm-project:mainfrom
wangx700:fix_ta_interface

wangx700 commented Mar 3, 2026 •

edited by github-actions bot

Loading

Uh oh!

gemini-code-assist bot commented Mar 3, 2026

Uh oh!

github-actions bot commented Mar 3, 2026

Uh oh!

gemini-code-assist bot left a comment

Uh oh!

gemini-code-assist bot Mar 3, 2026

Uh oh!

gemini-code-assist bot Mar 3, 2026

Uh oh!

gemini-code-assist bot Mar 3, 2026

Uh oh!

linfeng-yuan left a comment •

edited

Loading

Uh oh!

Ronald1995 Mar 4, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Conversation

wangx700 commented Mar 3, 2026 • edited by github-actions bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What this PR does / why we need it?

Does this PR introduce any user-facing change?

How was this patch tested?

Uh oh!

gemini-code-assist bot commented Mar 3, 2026

Summary of Changes

Highlights

Footnotes

Uh oh!

github-actions bot commented Mar 3, 2026

Uh oh!

gemini-code-assist bot left a comment

Choose a reason for hiding this comment

Code Review

Uh oh!

gemini-code-assist bot Mar 3, 2026

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist bot Mar 3, 2026

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist bot Mar 3, 2026

Choose a reason for hiding this comment

Uh oh!

linfeng-yuan left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Ronald1995 Mar 4, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

wangx700 commented Mar 3, 2026 •

edited by github-actions bot

Loading

linfeng-yuan left a comment •

edited

Loading