[Misc] ADD Docker compose exemple #20210
Conversation
|
👋 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 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 🚀 |
![gemini-code-assist[bot]](https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
There was a problem hiding this comment.
Summary of Changes
Hello @maher-naija-pro, 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 a new Docker Compose example to streamline the deployment and startup of the vLLM server. The provided configuration aims to simplify the process for users to quickly launch a vLLM instance with GPU support.
Highlights
- New Docker Compose Example: I've added a new file,
docker-compose.yml-exemple, which provides a complete Docker Compose configuration to easily set up and run the vLLM server. - vLLM Service Configuration: The example defines a
vllmservice using thevllm/vllm-openai:v0.9.1image, configured withnvidiaruntime for GPU access, port 8000 mapping, and a volume for models. - Customizable Server Parameters: The
commandsection within the Docker Compose file includes placeholders and examples for common vLLM server arguments such as--model,--tensor-parallel-size,--compilation-config, and--gpu-memory-utilization, along with an environment variable forHUGGING_FACE_HUB_TOKEN.
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
-
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. ↩
| ports: | ||
| - "8000:8000" | ||
| volumes: | ||
| - models:/models |
There was a problem hiding this comment.
The current volume configuration mounts a volume at /models, but vLLM and the underlying Hugging Face libraries download models to /root/.cache/huggingface by default. This means models won't be persisted in the models volume as intended. To correctly cache the models, you should mount the volume to /root/.cache/huggingface.
- models:/root/.cache/huggingface
| volumes: | ||
| - models:/models | ||
| environment: | ||
| HUGGING_FACE_HUB_TOKEN: "YOUR TOKEN HERE" |
There was a problem hiding this comment.
To ensure that models are stored in the correctly mounted volume, you should explicitly set the HF_HOME environment variable. This directs the Hugging Face library to use the specified path for caching.
HF_HOME: /root/.cache/huggingface
HUGGING_FACE_HUB_TOKEN: "YOUR TOKEN HERE"
| HUGGING_FACE_HUB_TOKEN: "YOUR TOKEN HERE" | ||
| restart: always | ||
| command: > | ||
| --model YOUR-MODEL-NALE |
There was a problem hiding this comment.
Typo: YOUR-MODEL-NALE should be YOUR-MODEL-NAME.
--model YOUR-MODEL-NAME
Essential Elements of an Effective PR Description Checklist
supported_models.mdandexamplesfor a new model.Purpose
Easy start vllm server with docker compose
Test Plan
NB: update tockn and model name on docker compose file
docker-compose up -d
Test Result
docker ps
=> vllm started
(Optional) Documentation Update
To start vllm with docker compose
Update env var with your token and model name
start docker compose
docker-compose up -d