[0.18.0][BugFix] Update capture sizes after rounding operations

[slippersss]

What this PR does / why we need it?

This PR is partially cherry-picked from #8172.

This PR aims to fix mismatched capture sizes after rounding operations when using sp or speculative. The reason is that original self.cudagraph_batch_sizes is no longer updated and remains as the initial sizes. Now we use self.cudagraph_dispatcher.get_capture_descs to the get up-to-date sizes.

Does this PR introduce any user-facing change?

N/A

How was this patch tested?

by ci

[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 addresses a bug where capture sizes became mismatched following rounding operations in speculative or sp modes. By shifting from static cached sizes to real-time descriptors provided by the dispatcher, the system ensures that graph parameters are correctly aligned with the actual state of the model runner.

Highlights

• Capture Size Synchronization: Updated the capture size retrieval mechanism to use the dispatcher's current descriptors instead of relying on stale cached values.
• Dynamic Graph Parameter Updates: Ensured that graph parameters for both standard and speculative configurations are initialized using the dynamically calculated capture sizes.

🧠 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. ↩

[gemini-code-assist]

Code Review

Suggested PR Title:\n\nmarkdown\n[Ops][Misc] Use dynamic capture sizes for ACL graph parameter initialization\n\n\nSuggested PR Summary:\n\nmarkdown\n### What this PR does / why we need it?\nThis pull request updates the `_check_and_update_cudagraph_mode` method to retrieve capture sizes from the `cudagraph_dispatcher`. This ensures that `set_graph_params` and `set_draft_graph_params` are initialized with the correct token counts for ACL graph execution instead of using static batch sizes. A review comment suggests simplifying the set comprehension used to extract these sizes for better readability.\n\n### Does this PR introduce _any_ user-facing change?\nNo.\n\n### How was this patch tested?\nCI passed with existing tests.\n

[gemini-code-assist]

The logic to extract capture_sizes from capture_descs is redundant and can be simplified using a set comprehension directly. This improves readability and maintainability.

capture_sizes = sorted({desc.num_tokens for descs in capture_descs.values() for desc in descs})

[yiz-liu]

It got a point, but not so important.