[perf] Batch rpyc calls in multimodal path -2

[SangChengC]

No description provided.

[gemini-code-assist]

Summary of Changes

Hello @SangChengC, 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 primarily focuses on optimizing the multimodal cache system by introducing batch processing for cache operations and enabling necessary RPyC serialization. These changes aim to reduce RPC overhead and improve the efficiency of data handling within the system, while also simplifying the underlying cache manager's design.

Highlights

• Performance Optimization: Batched Cache Operations: The cache manager's API has been refactored to support batch operations for allocating, releasing, and checking/setting item data and embeddings. This significantly reduces the number of individual RPyC calls, improving overall performance by minimizing RPC overhead.
• RPyC Configuration Update: All RPyC connections to the cache server now explicitly include config={'allow_pickle': True}. This enables the serialization and deserialization of complex Python objects, which is essential for passing data like torch.Tensor or custom data structures over the RPC connection.
• Cache Manager Architecture Simplification: The CacheManager interface and CacheManagerFactory abstraction have been removed. The InMemoryCache implementation is now directly used, simplifying the cache system's architecture and reducing indirection.

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. ↩

[gemini-code-assist]

Code Review

This pull request introduces batch rpyc calls in the multimodal path, aiming to enhance performance. However, it's crucial to address a logic error in the audio embedding cache, a race condition in the cache implementation, and an argument passing bug in the rpyc cache server.

[gemini-code-assist]

The calls to the underlying cache implementation (self._impl) are using keyword arguments (e.g., ids=ids), but the methods on InMemoryCache expect positional arguments. This will cause a TypeError at runtime. The arguments should be passed positionally.

Suggested change

	def exposed_set_items_data(self, ids: list[int]) -> None:
	ids = obtain(ids)
	return self._impl.set_items_data(ids=ids)
	def exposed_get_item_data(self, id: int) -> bool:
	id = obtain(id)
	return self._impl.get_item_data(id=id)
	def exposed_get_items_data(self, ids: list[int]) -> list[bool]:
	ids = obtain(ids)
	return self._impl.get_items_data(ids=ids)
	def exposed_set_item_embed(self, id: int) -> None:
	id = obtain(id)
	return self._impl.set_item_embed(id=id)
	def exposed_set_items_embed(self, ids: list[int]) -> None:
	ids = obtain(ids)
	return self._impl.set_items_embed(ids=ids)
	def exposed_get_item_embed(self, id: int) -> bool:
	id = obtain(id)
	return self._impl.get_item_embed(id=id)
	def exposed_get_items_embed(self, ids: list[int]) -> list[bool]:
	ids = obtain(ids)
	return self._impl.get_items_embed(ids=ids)
	def exposed_set_items_data(self, ids: list[int]) -> None:
	ids = obtain(ids)
	return self._impl.set_items_data(ids)
	def exposed_get_items_data(self, ids: list[int]) -> list[bool]:
	ids = obtain(ids)
	return self._impl.get_items_data(ids)
	def exposed_set_items_embed(self, ids: list[int]) -> None:
	ids = obtain(ids)
	return self._impl.set_items_embed(ids)
	def exposed_get_items_embed(self, ids: list[int]) -> list[bool]:
	ids = obtain(ids)
	return self._impl.get_items_embed(ids)