[Core] Add reload_weights RPC method

[22quinn]

Essential Elements of an Effective PR Description Checklist

• The purpose of the PR, such as "Fix some issue (link existing issues this PR will resolve)".
• The test plan, such as providing test command.
• The test results, such as pasting the results comparison before and after, or e2e results
• (Optional) The necessary documentation update, such as updating supported_models.md and examples for a new model.

Purpose

See #19886

Test Plan

See #19640

Test Result

See #19640

(Optional) Documentation Update

[gemini-code-assist]

Summary of Changes

Hello @22quinn, 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 introduces the core functionality for a new reload_weights RPC method, enabling the dynamic, in-place reloading of model weights for both GPU and TPU workers. This change refactors existing model loading logic to clearly separate initial model loading from subsequent weight reloading, and it includes improvements to memory pool context management.

Highlights

• Support for reload_weights RPC: Core reload_weights methods have been added to GPUModelRunner and TPUModelRunner to facilitate in-place reloading of model weights, laying the groundwork for an RPC method.
• Worker-level exposure: The new reload_weights functionality is exposed at the GPUWorker and TPUWorker levels, allowing higher-level components to trigger weight reloads.
• Refactored model loading: The load_model methods in GPUModelRunner and TPUModelRunner have been simplified to always perform a full model load, removing the previous conditional in-place weight reloading logic.
• Memory pool context helper: A new private helper method _maybe_get_memory_pool_context was introduced in GPUWorker to centralize and simplify the management of memory pool contexts during model and KV cache operations.

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 in your pull request via creating an issue comment (i.e. comment on the pull request page) using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands.

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 issue 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 is currently in preview and 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 to provide feedback.

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. ↩

[github-actions]

👋 Hi! Thank you for contributing to the vLLM project.

💬 Join our developer Slack at https://slack.vllm.ai to discuss your PR in #pr-reviews, coordinate on features in #feat- channels, or join special interest groups in #sig- channels.

Just a reminder: PRs would not trigger full CI run by default. Instead, it would only run fastcheck CI which starts running only a small and essential subset of CI tests to quickly catch errors. You can run other CI tests on top of those by going to your fastcheck build on Buildkite UI (linked in the PR checks section) and unblock them. If you do not have permission to unblock, ping simon-mo or khluu to add you in our Buildkite org.

Once the PR is approved and ready to go, your PR reviewer(s) can run CI to test the changes comprehensively before merging.

To run CI, PR reviewers can either: Add ready label to the PR or enable auto-merge.

🚀

[gemini-code-assist]

Code Review

This pull request introduces a reload_weights RPC method, which is a valuable addition for dynamically updating model weights. The implementation is clean and consistent across both GPU and TPU workers. The refactoring of load_model to separate concerns is a good design choice, and the consolidation of memory pool context logic in the GPU worker improves code quality.

My main feedback is on a potential side effect of the load_model refactoring in gpu_model_runner.py, which now always reloads the model from scratch. This changes its previous behavior and might affect existing callers or tests that expect an in-place weight reload. Please see the detailed comment on this.

[gemini-code-assist]

This change alters the behavior of load_model. Previously, if the model was already loaded (hasattr(self, "model")), it would reload the weights in-place. Now, load_model will always reload the entire model from scratch. While this simplifies load_model and the new reload_weights method correctly encapsulates the in-place loading, this is a significant behavior change.

This could be an issue if other parts of the codebase call load_model expecting it to only reload weights. For instance, the test test_load_model_weights_inplace in tests/v1/worker/test_gpu_model_runner.py appears to rely on the old behavior by calling load_model to perform what the test describes as an in-place weight load.

If this behavior change is intentional, I recommend updating the relevant tests to use the new reload_weights method to accurately reflect the new API design. Otherwise, the previous conditional logic should be restored to avoid breaking existing functionality.

[22quinn]

This is intentional and test is now updated

[mergify]

This pull request has merge conflicts that must be resolved before it can be
merged. Please rebase the PR, @22quinn.

https://docs.github.com/en/pull-requests/collaborating-with-pull-requests/working-with-forks/syncing-a-fork