[bugfix] Fix duplicate 'load_format' argument being passed in rollout

[hjh0119]

No description provided.

[gemini-code-assist]

Summary of Changes

Hello @hjh0119, 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 bug where the load_format argument could be inadvertently duplicated or mishandled during the initialization of the inference engine within the rollout process. The change refines how this argument is checked and defaulted, ensuring it is correctly applied at the top-level kwargs to prevent unexpected behavior and maintain the integrity of argument passing.

Highlights

• Bug Fix: Duplicate 'load_format' Argument: Corrected the logic within the get_infer_engine function to prevent the load_format argument from being duplicated or incorrectly handled when passed to the inference engine.
• Argument Handling Refinement: Ensured that the load_format argument is checked and defaulted to 'dummy' directly within the main kwargs dictionary, rather than engine_kwargs, to maintain proper argument flow.
• Code Clarity: Removed a potentially misleading comment related to the load_format argument's handling, improving code readability.

🧠 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.

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 aims to fix a bug with a duplicate load_format argument during rollout. While the change successfully resolves the TypeError crash, it appears to introduce a logic issue where the load_format is not correctly defaulted to 'dummy' for rollouts. My review includes a suggestion to make the logic more robust to ensure the correct behavior for RL weight synchronization.

[gemini-code-assist]

This change correctly fixes the TypeError from a duplicate load_format argument. However, it introduces a potential logic bug. args.get_vllm_engine_kwargs() likely provides a default value of 'auto' for load_format, which is then added to kwargs. In that case, the condition 'load_format' not in kwargs will be false, and the intended default of 'dummy' for rollouts will not be set. This could lead to incorrect behavior during RL rollout, as vLLM might try to load a model from a path named 'auto' instead of waiting for weight synchronization. A more robust check is needed to handle the default 'auto' value.

# For RL rollout, we use 'dummy' load_format to prevent vLLM from loading weights from disk,
# as they will be synced from the trainer process.
# We override the default 'auto' but respect other user-provided values.
if kwargs.get('load_format') is None or kwargs.get('load_format') == 'auto':
    kwargs['load_format'] = 'dummy'