Improve vLLM examples regarding vllm_engine_kwargs use #7133
Conversation
Summary of ChangesHello @3manifold, 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 the existing vLLM examples by introducing a practical demonstration of how to utilize the Highlights
🧠 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 AssistThe 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
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 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
|
![gemini-code-assist[bot]](https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
There was a problem hiding this comment.
Code Review
This pull request adds a valuable example demonstrating the use of --vllm_engine_kwargs to configure the attention backend in vLLM. The change is correct and enhances the provided examples. I've added one suggestion to include a comment about the required vLLM version to make the example even more user-friendly.
| --val_dataset speech_asr/speech_asr_aishell1_trainsets:validation#1000 \ | ||
| --vllm_gpu_memory_utilization 0.9 \ | ||
| --vllm_max_model_len 8192 \ | ||
| --vllm_engine_kwargs '{"attention_config": {"backend": "FLASH_ATTN"}}' \ |
There was a problem hiding this comment.
This is a great example of using vllm_engine_kwargs. To prevent potential issues for users with older vLLM versions, it would be helpful to add a comment specifying the version requirement for attention_config, as this feature was introduced in vllm>=0.3.0.
| --vllm_engine_kwargs '{"attention_config": {"backend": "FLASH_ATTN"}}' \ | |
| # Use flash attention via vllm_engine_kwargs (requires vllm>=0.3.0) | |
| --vllm_engine_kwargs '{"attention_config": {"backend": "FLASH_ATTN"}}' \ |
1 participant
PR type
PR information
Add an example of an advanced usecase of
--vllm_engine_kwargsuse since there was nothing similar in the past. It applies tovllm >= 0.13.0.List of input vLLM args can be found here vllm-project/vllm#26315 . It relates to #7132 .