Repository navigation
[CI][Bugfix] Initialize _transfer_layer_group_ids in region_pull_worker fixture - #57211
Merged
robertgshaw2-redhat merged 2 commits intoSep 16, 2026
Conversation
…er fixture PR vllm-project#57077's region_pull_worker fixture constructs NixlConnectorWorker via object.__new__() and sets attributes manually. It does not set _transfer_layer_group_ids, which base_worker.py:198 reads (self._transfer_layer_group_ids) since vllm-project#50494 merged. vllm-project#57077's CI ran against a base without vllm-project#50494, so the conflict only surfaces on main: tests/v1/kv_connector/unit/test_nixl_connector_hma.py::test_region_pull_ignores_allocation_padding fails 7x with AttributeError (build #89376). Set the attribute to the empty-tuple default used by the base class __init__ (and by the sibling _make_mock_worker_for_desc_ids fixture). Signed-off-by: kimi-no-na-wa <kimi-no-na-wa@users.noreply.github.com>
vllm-agent
force-pushed
the
fix-nixl-hma-transfer-layer-group-ids
branch
from
September 16, 2026 18:22
b24e491 to
5d65955
Compare
Contributor
|
@ApostaC @orozery @ivanium @NickLucche @xuechendi mind taking a look? This is red on main (reproduced on a clean 403f182) and failing unrelated PRs including mine (#57070). One-line fixture fix, but it needs write access to un-draft it and start CI |
Collaborator
Author
|
/ci run |
|
❌ @vllm-agent, PR authors cannot run CI while the PR is a draft. |
Contributor
|
@vllm-agent Please open this as a PR, not a draft. |
robertgshaw2-redhat
marked this pull request as ready for review
September 16, 2026 20:05
robertgshaw2-redhat
requested review from
ApostaC,
ivanium and
orozery
as code owners
September 16, 2026 20:05
![claude[bot]](https://avatars.githubusercontent.com/in/1236702?s=60&v=4){kind=small}
robertgshaw2-redhat
added
the
ready
Collaborator
|
/ci run |
|
✅ @vllm-agent, CI is now available for this PR.
|
|
✅ Triggered Buildkite CI #89432 for commit |
robertgshaw2-redhat
approved these changes
Sep 16, 2026
robertgshaw2-redhat
enabled auto-merge (squash)
September 16, 2026 20:06
matteso1
added a commit
to matteso1/vllm
that referenced
this pull request
Sep 16, 2026
Carry the one-line fix from vllm-project#57211 (5d65955). Initialize the constructor-bypassing fixture with the production empty-tuple default; transfer behavior and test assertions are unchanged. Signed-off-by: Nils Matteson <nilsmatteson@icloud.com> (cherry picked from commit bf4e6c1)
matteso1
added a commit
to matteso1/vllm
that referenced
this pull request
Sep 16, 2026
Carry the one-line fix from vllm-project#57211 (5d65955). Initialize the constructor-bypassing fixture with the production empty-tuple default; transfer behavior and test assertions are unchanged. Signed-off-by: Nils Matteson <nilsmatteson@icloud.com> (cherry picked from commit bf4e6c1)
matteso1
added a commit
to thaw-ai/vllm
that referenced
this pull request
Sep 16, 2026
Carry the one-line fix from vllm-project#57211 (5d65955). Initialize the constructor-bypassing fixture with the production empty-tuple default; transfer behavior and test assertions are unchanged. Signed-off-by: Nils Matteson <nilsmatteson@icloud.com>
NickLucche
approved these changes
Sep 16, 2026
michael-gpu
pushed a commit
to practical-inference/vllm-ampere-extended
that referenced
this pull request
Oct 20, 2026
…er fixture (vllm-project#57211) Signed-off-by: kimi-no-na-wa <kimi-no-na-wa@users.noreply.github.com> Co-authored-by: kimi-no-na-wa <kimi-no-na-wa@users.noreply.github.com> Co-authored-by: Robert Shaw <114415538+robertgshaw2-redhat@users.noreply.github.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Problem
Main is red on
:computer: (CPU) V1 Others(builds #89375, #89376): all 7 parametrizations oftest_region_pull_ignores_allocation_paddingfail withRoot cause is a stale-base merge conflict between two PRs:
_transfer_layer_group_idsattribute to the NIXL base worker and reads it inbase_worker.py:198(use_layer_name_routing = bool(self._transfer_layer_group_ids)).region_pull_workerfixture, which constructs the worker viaobject.__new__(NixlConnectorWorker)and sets attributes manually — without_transfer_layer_group_ids. Its CI ran against a base that did not yet contain [KVConnector][NIXL] Support attention-HMA layouts in pipeline-parallel push prefill #50494, so the conflict only surfaced after both landed on main.Fix
Set
worker._transfer_layer_group_ids = ()in the fixture — the same empty-tuple default the base-class__init__uses, and the same value the sibling_make_mock_worker_for_desc_idsfixture already sets.Note: not runnable locally on my box (no torch), so validation is via CI — pre-commit hooks (ruff, mypy, DCO) pass.
Signed-off-by: kimi-no-na-wa kimi-no-na-wa@users.noreply.github.com