[BugFix][310p] Align 310P GDN state semantics with vLLM

[Tflowers-0129]

What this PR does / why we need it?

• I found that the FLA state semantics are not aligned between vllm-ascend and vLLM: in vLLM, the state layout is [N, HV, V, K], while in vllm-ascend it is [N, HV, K, V]. The early 310P implementation was aligned with the mainline vllm-ascend behavior, but the operator team now plans to provide a fused recurrent-gated-delta-rule operator whose semantics will align with vLLM. Therefore, this needs to be adjusted in advance.

• I also found that L2 normalization can use the operator provided in functional, instead of being fully implemented through manually stitched small ops as it is now.

Does this PR introduce any user-facing change?

NA

How was this patch tested?

ut and the e2e test

• vLLM version: v0.18.0
• vLLM main: vllm-project/vllm@35141a7

[gemini-code-assist]

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 aligns the 310P GDN state semantics with vLLM by adjusting the state layout and refactoring L2 normalization to use standard library functions. These changes ensure consistency across the codebase and prepare for upcoming operator updates.

Highlights

• State Layout Alignment: Updated the FLA state semantics layout from [N, HV, K, V] to [N, HV, V, K] to ensure compatibility with vLLM's expected format.
• L2 Normalization Optimization: Replaced manual L2 normalization implementations with PyTorch's built-in functional operator for improved efficiency and maintainability.

🧠 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. ↩

[github-actions]

👋 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

This PR aligns the state layout of Gated Delta Rule (GDN) operations with vLLM by swapping the K and V dimensions in state tensors. Feedback focuses on maintaining numerical stability by avoiding F.normalize in favor of an explicit L2 norm with a larger epsilon for float16 compatibility, and correcting the PR title format.

Suggested PR Title:

[310p][Ops][BugFix] Align 310P GDN state semantics with vLLM

Suggested PR Summary:

### What this PR does / why we need it?
This PR aligns the state layout of Gated Delta Rule (GDN) operations (chunked and fused recurrent) with vLLM's expected semantics by changing state shapes from `[..., K, V]` to `[..., V, K]`. This ensures consistency with reference FLA kernels.

Feedback was provided to revert the change from manual L2 normalization to `torch.nn.functional.normalize`, as the latter's default epsilon (1e-12) is too small for `float16` stability.

### Does this PR introduce _any_ user-facing change?
Yes, the expected shape of the `initial_state` tensor for GDN operations has changed.

### How was this patch tested?
Tested via updated unit tests and new parity tests verifying state layout against vLLM reference values.