[Bugfix] qwen2.5-vl spec decode accept_len low

[Lzhang-hub]

Motivation

After this pr, qwen2.5-vl eagle3 infer draft accept_len is low.

Because aux_hidden_states is miss in logits_processor.

Releated issue: sgl-project/SpecForge#310

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.
• Work with maintainers to merge your PR. See the PR Merge Process

[gemini-code-assist]

Summary of Changes

Hello @Lzhang-hub, 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 a performance regression in the qwen2.5-vl model's speculative decoding mechanism. By correctly passing a previously omitted parameter, aux_hidden_states, within the model's forward pass, the change aims to restore the expected accept_len and improve inference efficiency for this model.

Highlights

• Bugfix: qwen2.5-vl accept_len: This pull request addresses a bug where the accept_len for the qwen2.5-vl model was unexpectedly low after a previous update. The issue was traced to the omission of aux_hidden_states in the logits_processor.
• Code Correction: The fix involves re-introducing the aux_hidden_states parameter to the relevant function call within the forward method of the qwen2_5_vl model, ensuring correct speculative decoding behavior.

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.

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

This pull request addresses a bug in the speculative decoding implementation for the qwen2.5-vl model, where the accept_len was low. The root cause was the missing aux_hidden_states argument in the call to the logits_processor. This change correctly passes aux_hidden_states, which is essential for the EAGLE3 speculative decoding algorithm. The fix is accurate, minimal, and directly addresses the issue described. The code is now consistent with other models supporting this feature. I find the change to be correct and have no further suggestions.

[yhyang201]

cc @yuan-luo

[yhyang201]

/tag-and-rerun-ci

[yuan-luo]

Thanks for the fix. LGTM.

[yhyang201]

@mickqian I think this is ready to be merged. Could you please merge it?