updates for vLLM compativility with lora

[danielhanchen]

No description provided.

[gemini-code-assist]

Summary of Changes

Hello @danielhanchen, 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 key updates to enhance the compatibility of LoRA (Low-Rank Adaptation) with vLLM. The changes focus on refining the process of loading LoRA adapter tensors and embeddings, ensuring a more unified approach. Additionally, a previous constraint on the LoRA's extra vocabulary size has been relaxed, which should improve the flexibility and robustness of integrating various LoRA configurations within the vLLM framework.

Highlights

• LoRA Tensor and Embedding Consolidation: The _load_adapter method now combines lora_tensors and lora_embeddings into a single dictionary before passing them to the LoRA model constructor, streamlining the data handling for adapter loading.
• Robust Configuration Access: The calculation for target_embedding_padding has been updated to use getattr when accessing lora_extra_vocab_size from the configuration, making the code more resilient to cases where this attribute might be missing.
• Removed Extra Vocab Size Validation: A previous validation check that raised an error if the LoRA's extra_vocab_size exceeded the configured lora_extra_vocab_size has been removed, potentially allowing for greater flexibility in LoRA model integration.

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 updates the LoRA loading mechanism for compatibility with vLLM. The key changes include merging LoRA tensors and embeddings into a single dictionary and making the lora_extra_vocab_size configuration optional for more flexibility. These changes appear to align with updates in vLLM. However, I have pointed out a potential issue with the removal of a validation check for the extra vocabulary size, which could lead to silent failures if not handled elsewhere. I've suggested reintroducing a more robust version of this check to maintain code safety.

[gemini-code-assist]

The validation check for lora.extra_vocab_size has been removed. While making lora_extra_vocab_size optional in lora_config is a good improvement for flexibility, removing this validation entirely could be risky. If a LoRA adapter adds new vocabulary and lora_extra_vocab_size is not configured (defaulting to 0), it might lead to silent errors or incorrect model behavior if the underlying LoRAModel creation methods don't perform this validation.

It's safer to ensure this validation happens. If it's not handled elsewhere (e.g., within LoRAModel creation or peft_helper.validate_legal), please consider reintroducing a more robust version of this check.

        lora_extra_vocab_size = getattr(self.lora_config, "lora_extra_vocab_size", 0)
        if lora.extra_vocab_size > lora_extra_vocab_size:
            raise ValueError(
                f"LoRA added vocab size {lora.extra_vocab_size} is greater "
                f"than lora_extra_vocab_size {lora_extra_vocab_size}."
            )
        return lora