[add] add skip image cache and disable_prompt_cache para

[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 enhances control over caching mechanisms within the system by introducing new parameters for both text prompt and multimodal image processing. It adds disable_prompt_cache to SamplingParams and skip_image_cache to MultimodalParams, providing users with the flexibility to explicitly bypass caching for specific requests. These changes are integrated into the core inference and visual server components, alongside a minor refinement to shared memory size calculation for multimodal inputs.

Highlights

• New Parameter: disable_prompt_cache: Introduced a new boolean parameter, disable_prompt_cache, within SamplingParams to allow explicit control over whether prompt caching should be bypassed for a given request. This parameter is integrated into the inference batch processing logic.
• New Parameter: skip_image_cache: Added a new boolean parameter, skip_image_cache, to MultimodalParams. This enables requests to bypass the image embedding cache, ensuring that image embeddings are re-processed rather than retrieved from cache.
• Prompt Cache Management: Modified the inference batch processing to respect the disable_prompt_cache parameter, preventing prompt cache matching and memory freeing for requests where caching is explicitly disabled.
• Image Cache Management: Updated the visual server manager to check the skip_image_cache parameter for multimodal requests, allowing it to bypass the image embedding cache lookup when set to true.
• Shared Memory Size Calculation: Adjusted the shared memory size estimation utility by adding image_patch_max_num to fake image items, which helps in more accurately calculating the required shared memory for multimodal image tokens.

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 introduces two new parameters, skip_image_cache and disable_prompt_cache, to provide more control over caching mechanisms. The implementation correctly integrates these parameters into the multimodal parameter handling and the inference batch processing logic. The changes are logical and well-contained. I've provided a few minor suggestions to improve code quality, including correcting a comment, removing dead code, and refactoring a magic number into a constant for better maintainability.

[gemini-code-assist]

The comment for disable_prompt_cache appears to be a copy-paste from the line above and is incorrect. It should be updated to accurately describe the purpose of this parameter.

Suggested change

	("disable_prompt_cache", ctypes.c_bool), # eos_id will be always ignored except the value is set to True
	("disable_prompt_cache", ctypes.c_bool), # whether to disable prompt cache

[gemini-code-assist]

The value 12 is a magic number. It should be defined as a named constant at the module level (e.g., DEFAULT_IMAGE_PATCH_MAX_NUM = 12) for better readability and maintainability.

Suggested change

	fake_image_item.extra_params["image_patch_max_num"] = 12
	fake_image_item.extra_params["image_patch_max_num"] = DEFAULT_IMAGE_PATCH_MAX_NUM