Skip to content

[NIXL] vllm v0 nixl integration #16677

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Draft
wants to merge 7 commits into
base: main
Choose a base branch
from

Conversation

rainj-me
Copy link

@rainj-me rainj-me commented Apr 15, 2025

Features are supported in this PR

  • vllm v0 nixl integration without etcd/nats as dependencies
  • http apis to handshake nixl conns
  • MLA layout support
  • MTP and speculative decoding support
  • Remote prefill worker skip sampling to reduce TTFT

Usage scripts

# Decode worker
vllm serve /data/models/QwQ-32B --tensor_parallel-size 2 --port 8080 --max-model-len 131072 --swap-space 0  --block-size 128 --trust-remote-code --kv-transfer-config '{"kv_connector":"DynamoNixlConnector"}' --enable-chunked-prefill false

# Remote prefill worker
vllm serve /data/models/QwQ-32B --tensor_parallel-size 1 --port 8090 --max-model-len 65536 --swap-space 0  --block-size 128 --gpu-memory-utilization 0.95 --trust-remote-code --enforce-eager --kv-transfer-config '{"kv_connector":"DynamoNixlConnector", "kv_connector_extra_config":{"skip_sampling":true}}' --enable-chunked-prefill false

# Establish nixl conn
curl -kvvv -XPOST http://127.0.0.1:8080/add_remote_prefill_eps  -H "Content-Type: application/json" -d '{"endpoints":["http://127.0.0.1:8090"]}'

# Test command
curl http://127.0.0.1:8080/v1/chat/completions -H "Content-Type: application/json" -d '{"model": "/data/models/QwQ-32B", "temperature": 0, "messages": [ {"role": "user", "content": "San Francisco is "}], "max_tokens": 300, "stream": false}'

What are left

  1. chunked-prefill support
  2. PP support
  3. vllm V1 support
  4. Only keep DynamoNixlConnector and cleanup other connectors

Copy link

👋 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 fastcheck CI which starts running only a small and essential subset of CI tests to quickly catch errors. You can run other CI tests on top of those by going to your fastcheck build on Buildkite UI (linked in the PR checks section) and unblock them. If you do not have permission to unblock, ping simon-mo or khluu to add you in our Buildkite org.

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 ready label to the PR or enable auto-merge.

🚀

Copy link

mergify bot commented Apr 15, 2025

This pull request has merge conflicts that must be resolved before it can be
merged. Please rebase the PR, @rainj-me.

https://docs.github.com/en/pull-requests/collaborating-with-pull-requests/working-with-forks/syncing-a-fork

@mergify mergify bot added the needs-rebase label Apr 15, 2025
Copy link

mergify bot commented Apr 19, 2025

This pull request has merge conflicts that must be resolved before it can be
merged. Please rebase the PR, @rainj-me.

https://docs.github.com/en/pull-requests/collaborating-with-pull-requests/working-with-forks/syncing-a-fork

@mergify mergify bot added the needs-rebase label Apr 19, 2025
Signed-off-by: Changqi Lu <luchangqi.123@bytedance.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants