Revert "Fix potential eos bug on decode instance when PD is enabled"

[hnyls2002]

Reverts #12206 @whybeyoung

[gemini-code-assist]

Summary of Changes

Hello @hnyls2002, 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 undoes a previous modification intended to fix an EOS-related issue within the decoding process, particularly when a disaggregated architecture is in use. By reverting the change, the system's behavior regarding when a request is considered immediately finished is restored to its state before the problematic fix, suggesting the original 'fix' may have introduced new issues or was not the correct solution for the identified bug.

Highlights

• Revert Previous Fix: This pull request reverts a prior change (Fix potential eos bug on decode instance when PD is enabled #12206) that aimed to address a potential end-of-sequence (EOS) bug in the decode instance when PD (presumably PagedAttention or similar disaggregation) is enabled.
• EOS Token Handling: The reversion specifically removes the logic that would cause a decode request to finish immediately if an EOS token was generated. The immediate finishing condition now solely relies on sampling_params.max_new_tokens == 1.

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 reverts a change that introduced a fast path for handling EOS tokens. The revert is correct as the original implementation did not account for the ignore_eos sampling parameter, potentially causing incorrect behavior. I've suggested a way to re-introduce the optimization safely, which you may consider for a future change.

[gemini-code-assist]

This revert is correct because the original change didn't handle the ignore_eos sampling parameter, which could cause premature termination of a request. While reverting is a safe fix, the optimization for early EOS exit can be re-introduced by also checking for ignore_eos.

                if (
                    decode_req.req.sampling_params.max_new_tokens == 1
                    or (
                        output_id in decode_req.req.eos_token_ids
                        and not decode_req.req.sampling_params.ignore_eos
                    )
                ):