Skip to content

Conversation

@xuechendi
Copy link
Contributor

@xuechendi xuechendi commented Sep 18, 2025

Purpose

Originally, nixl_connector only supports in-tree platforms and UCX as backend. In this PR, we want to extend ths nixl_connector support to OOT platform as well. Also, want to introduce the option to use different NIXL_backend instead of UCX for vllm

Test Plan

use existing test

Test Result


Essential Elements of an Effective PR Description Checklist
  • The purpose of the PR, such as "Fix some issue (link existing issues this PR will resolve)".
  • The test plan, such as providing test command.
  • The test results, such as pasting the results comparison before and after, or e2e results
  • (Optional) The necessary documentation update, such as updating supported_models.md and examples for a new model.
  • (Optional) Release notes update. If your change is user facing, please update the release notes draft in the Google Doc.

Copy link
Contributor

@gemini-code-assist gemini-code-assist bot left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Code Review

This pull request extends the nixl_connector to support Out-Of-Tree (OOT) platforms and allows configuring the NIXL backend via an environment variable. The changes are well-structured, following the existing platform interface pattern in vLLM by adding new methods to the Platform class. This makes the integration clean and extensible. I have one suggestion to improve the robustness of handling the new environment variable.

@xuechendi
Copy link
Contributor Author

xuechendi commented Sep 18, 2025

@robertgshaw2-redhat @NickLucche please help to review this PR, thanks

Copy link
Collaborator

@NickLucche NickLucche left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Hey @xuechendi , thanks a lot for contributing!

This looks clean, but I also agree with @chaunceyjiang it would be even cleaner to put extra configs in kv_connector_extra_config":{"backend":"ucx"}} instead of adding more env vars.

Also, it would be quite useful to the community if you could add a line or two to the docs highlighting your change, as well as a simple unit test to test_nixl_connector.py checking memory_type/kv_buffer_type are as intended (we have more changes lined up to it in review).
Thank you!

@mergify mergify bot added the kv-connector label Sep 18, 2025
@xuechendi xuechendi requested a review from hmellor as a code owner September 19, 2025 01:26
@mergify mergify bot added the documentation Improvements or additions to documentation label Sep 19, 2025
@xuechendi
Copy link
Contributor Author

xuechendi commented Sep 19, 2025

@NickLucche, for adding a new UT in test_nixl_connector.py, I added one "test_kv_buffer_to_nixl_memory_types", which is initialized with a FakePlatform, may you check if this is what you expected?

@mergify mergify bot added the v1 label Sep 19, 2025
@mergify mergify bot added the ci/build label Sep 19, 2025
@@ -63,6 +64,8 @@
"tpu": ("cpu", ),
"xpu": ("cpu", ),
}
# support for oot platform by providing mapping in current_platform
_NIXL_SUPPORTED_DEVICE.update(current_platform.get_nixl_supported_devices())
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The variable _NIXL_SUPPORTED_DEVICE appears to be unused.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@mergify
Copy link

mergify bot commented Sep 19, 2025

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

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

@xuechendi
Copy link
Contributor Author

@NickLucche @chaunceyjiang , may you help to check again, or please approve if the PR is looking good to you, Thanks

Copy link
Collaborator

@NickLucche NickLucche left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM, thanks @xuechendi !

@NickLucche NickLucche enabled auto-merge (squash) September 22, 2025 16:49
@github-actions github-actions bot added the ready ONLY add when PR is ready to merge/full CI is needed label Sep 22, 2025
auto-merge was automatically disabled September 22, 2025 19:38

Head branch was pushed to by a user without write access

Signed-off-by: Chendi Xue <Chendi.Xue@intel.com>
Signed-off-by: Chendi Xue <Chendi.Xue@intel.com>
Signed-off-by: Chendi Xue <Chendi.Xue@intel.com>
Signed-off-by: Chendi Xue <Chendi.Xue@intel.com>
Signed-off-by: Chendi Xue <Chendi.Xue@intel.com>
Signed-off-by: Chendi Xue <Chendi.Xue@intel.com>
Signed-off-by: Chendi Xue <Chendi.Xue@intel.com>
@xuechendi xuechendi force-pushed the dev/nixl_plugin branch 2 times, most recently from ae425a2 to 6e21b38 Compare September 22, 2025 23:01
@jikunshang jikunshang enabled auto-merge (squash) September 23, 2025 00:00
Signed-off-by: Chendi Xue <Chendi.Xue@intel.com>
auto-merge was automatically disabled September 23, 2025 00:43

Head branch was pushed to by a user without write access

@jikunshang jikunshang enabled auto-merge (squash) September 23, 2025 00:49
Signed-off-by: Chendi Xue <Chendi.Xue@intel.com>
auto-merge was automatically disabled September 23, 2025 02:03

Head branch was pushed to by a user without write access

@jikunshang jikunshang enabled auto-merge (squash) September 23, 2025 03:17
@jikunshang jikunshang merged commit 5774b0a into vllm-project:main Sep 23, 2025
46 checks passed
FeiDaLI pushed a commit to FeiDaLI/vllm that referenced this pull request Sep 25, 2025
…er nixl_backend (vllm-project#25121)

Signed-off-by: Chendi Xue <Chendi.Xue@intel.com>
charlifu pushed a commit to ROCm/vllm that referenced this pull request Sep 25, 2025
…er nixl_backend (vllm-project#25121)

Signed-off-by: Chendi Xue <Chendi.Xue@intel.com>
Signed-off-by: charlifu <charlifu@amd.com>
yewentao256 pushed a commit that referenced this pull request Oct 3, 2025
…er nixl_backend (#25121)

Signed-off-by: Chendi Xue <Chendi.Xue@intel.com>
Signed-off-by: yewentao256 <zhyanwentao@126.com>
gjc0824 pushed a commit to gjc0824/vllm that referenced this pull request Oct 10, 2025
…er nixl_backend (vllm-project#25121)

Signed-off-by: Chendi Xue <Chendi.Xue@intel.com>
Signed-off-by: gaojc <1055866782@qq.com>
xuebwang-amd pushed a commit to xuebwang-amd/vllm that referenced this pull request Oct 10, 2025
…er nixl_backend (vllm-project#25121)

Signed-off-by: Chendi Xue <Chendi.Xue@intel.com>
Signed-off-by: xuebwang-amd <xuebwang@amd.com>
choprahetarth pushed a commit to Tandemn-Labs/vllm that referenced this pull request Oct 11, 2025
…er nixl_backend (vllm-project#25121)

Signed-off-by: Chendi Xue <Chendi.Xue@intel.com>
lywa1998 pushed a commit to lywa1998/vllm that referenced this pull request Oct 20, 2025
…er nixl_backend (vllm-project#25121)

Signed-off-by: Chendi Xue <Chendi.Xue@intel.com>
xuebwang-amd pushed a commit to xuebwang-amd/vllm that referenced this pull request Oct 24, 2025
…er nixl_backend (vllm-project#25121)

Signed-off-by: Chendi Xue <Chendi.Xue@intel.com>
Signed-off-by: xuebwang-amd <xuebwang@amd.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

ci/build documentation Improvements or additions to documentation kv-connector ready ONLY add when PR is ready to merge/full CI is needed v1

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants