Skip to content

Add ucx test#711

Merged
xuechendi merged 14 commits intovllm-project:mainfrom
pi314ever:ucx-nixl-ci
Jan 10, 2026
Merged

Add ucx test#711
xuechendi merged 14 commits intovllm-project:mainfrom
pi314ever:ucx-nixl-ci

Conversation

@pi314ever
Copy link
Copy Markdown
Contributor

@pi314ever pi314ever commented Dec 12, 2025 •
edited
Loading

Adds pd nixl UCX test to CI. This is dependent on intel-staging/ucx#1

Copilot AI review requested due to automatic review settings December 12, 2025 01:15
@github-actions
Copy link
Copy Markdown

github-actions bot commented Dec 12, 2025

🚧 CI Blocked

The main CI workflow was not started for the following reason:

Your branch is behind the base branch. Please merge or rebase to get the latest changes.

Copilot AI reviewed Dec 12, 2025
View reviewed changes
Copy link
Copy Markdown
Contributor

Copilot AI left a comment

Choose a reason for hiding this comment

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

Pull request overview

This PR adds support for UCX testing in the CI pipeline by introducing a new test function and making the UCX transport layer selection configurable based on the buffer device type.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 2 comments.

File Description
tests/unit_tests/run_accuracy_test.sh Makes UCX_TLS configurable, using gaudi-specific settings when buffer device is HPU, and restores proper git root detection
tests/full_tests/ci_gsm8k_tests.sh Adds new test function for PD disaggregate through NIXL UCX

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

@github-actions github-actions bot mentioned this pull request Dec 12, 2025
Open
@xuechendi xuechendi self-assigned this Dec 12, 2025
@xuechendi xuechendi marked this pull request as draft December 12, 2025 15:50
@xuechendi
Copy link
Copy Markdown
Collaborator

xuechendi commented Dec 12, 2025

Since intel-staging/ucx#1 is not merged yet, convert the PR to draft to avoid repeated CI
@pi314ever , once the UCX PR merged, please convert back to ready for review and update branch to retrigger CI

@pi314ever pi314ever mentioned this pull request Dec 13, 2025
Draft
@github-actions
Copy link
Copy Markdown

github-actions bot commented Jan 6, 2026

🚧 CI Blocked

The main CI workflow was not started for the following reason:

This is a Draft PR. Please mark it as 'Ready for Review' to trigger the CI.

@github-actions
Copy link
Copy Markdown

github-actions bot commented Jan 6, 2026

🚧 CI Blocked

The main CI workflow was not started for the following reason:

This is a Draft PR. Please mark it as 'Ready for Review' to trigger the CI.

@pi314ever pi314ever marked this pull request as ready for review January 6, 2026 00:04
@pi314ever pi314ever mentioned this pull request Jan 9, 2026
Closed
@github-actions
Copy link
Copy Markdown

github-actions bot commented Jan 9, 2026

🚧 CI Blocked

The main CI workflow was not started for the following reason:

Your branch is behind the base branch. Please merge or rebase to get the latest changes.

@pi314ever
Add ucx test
d4dc568 
Signed-off-by: Daniel Huang <daniel1.huang@intel.com>
@pi314ever
Add install script
c4a3a9e 
Signed-off-by: Daniel Huang <daniel1.huang@intel.com>
@pi314ever
gaudi_gdr ucx plugin support in install nixl
0a6521f 
Signed-off-by: Daniel Huang <daniel1.huang@intel.com>
@pi314ever
Add comment on ucx commit
f7847e8 
Signed-off-by: Daniel Huang <daniel1.huang@intel.com>
@pi314ever
Bring in functions directly for more robust plugin bundling
87eec88 
Signed-off-by: Daniel Huang <daniel1.huang@intel.com>
@pi314ever
Logging and cleanup
59a54d2 
Signed-off-by: Daniel Huang <daniel1.huang@intel.com>
@pi314ever
Update all system run commands to run_command
3c2f1db 
Signed-off-by: Daniel Huang <daniel1.huang@intel.com>
@pi314ever
Minor comment on where plugin bundle wheel functions come from
2939df6 
Signed-off-by: Daniel Huang <daniel1.huang@intel.com>
@pi314ever
Fix pre-commit checks
3990831 
Signed-off-by: Daniel Huang <daniel1.huang@intel.com>
@pi314ever
Switch to install_nixl.py upgrade
9ace647 
Signed-off-by: Daniel Huang <daniel1.huang@intel.com>
@pi314ever
Add wheels cache
17fe91b 
Signed-off-by: Daniel Huang <daniel1.huang@intel.com>
@pi314ever
Fix cwd for ucx acc test
5d4b85d 
Signed-off-by: Daniel Huang <daniel1.huang@intel.com>
@pi314ever
Fix device_to_device var and add ucx_memtype_cache env var
608b967 
Signed-off-by: Daniel Huang <daniel1.huang@intel.com>
xuechendi
xuechendi reviewed Jan 9, 2026
View reviewed changes
tests/unit_tests/run_accuracy_test.sh
export VLLM_NIXL_DEVICE_TO_DEVICE=true
UCX_TLS="tcp"
if [ "$VLLM_NIXL_BACKEND" == "UCX" ]; then
export VLLM_NIXL_DEVICE_TO_DEVICE=false
Copy link
Copy Markdown
Collaborator

@xuechendi xuechendi Jan 9, 2026

Choose a reason for hiding this comment

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

we will also test UCX + CPU with non_gaudi_gdr, so please don't put
VLLM_NIXL_DEVICE_TO_DEVICE=true as only option

Copy link
Copy Markdown
Contributor Author

@pi314ever pi314ever Jan 9, 2026

Choose a reason for hiding this comment

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

Fixed, added back VLLM_NIXL_DEVICE_TO_DEVICE=false for UCX + CPU configuration.

@pi314ever
Allow device_to_device=false for ucx+cpu use case
ad6cfd4 
Signed-off-by: Daniel Huang <daniel1.huang@intel.com>
@github-actions
Copy link
Copy Markdown

github-actions bot commented Jan 10, 2026

✅ CI Passed

All checks passed successfully against the following vllm commit:
aa125ecf0edb9cd67656553d11d643aeb444ff9e

@xuechendi xuechendi merged commit c885554 into vllm-project:main Jan 10, 2026
52 checks passed
jinyouzhi pushed a commit to jinyouzhi/vllm-gaudi that referenced this pull request Jan 14, 2026
@pi314ever @jinyouzhi
Add ucx test (vllm-project#711)
591dd9e 
Adds pd nixl UCX test to CI. This is dependent on
intel-staging/ucx#1

---------

Signed-off-by: Daniel Huang <daniel1.huang@intel.com>
Signed-off-by: Jin, Youzhi <youzhi.jin@intel.com>
slokesha pushed a commit to libinta/vllm-gaudi that referenced this pull request Feb 9, 2026
@pi314ever @slokesha
Add ucx test (vllm-project#711)
eaeba25 
Adds pd nixl UCX test to CI. This is dependent on
intel-staging/ucx#1

---------

Signed-off-by: Daniel Huang <daniel1.huang@intel.com>
Signed-off-by: slokesha <slokeshappa@habana.ai>
rajanintel24 pushed a commit to rajanintel24/vllm-gaudi that referenced this pull request Feb 11, 2026
@pi314ever @rajanintel24
Add ucx test (vllm-project#711)
12609bb 
Adds pd nixl UCX test to CI. This is dependent on
intel-staging/ucx#1

---------

Signed-off-by: Daniel Huang <daniel1.huang@intel.com>
adobrzyn pushed a commit that referenced this pull request Mar 31, 2026
@pi314ever @adobrzyn
Add ucx test (#711)
f5fa7b9 
Adds pd nixl UCX test to CI. This is dependent on
intel-staging/ucx#1

---------

Signed-off-by: Daniel Huang <daniel1.huang@intel.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

Copilot code review Copilot Copilot left review comments

@xuechendi xuechendi xuechendi approved these changes

@kzawora-intel kzawora-intel Awaiting requested review from kzawora-intel

@adobrzyn adobrzyn Awaiting requested review from adobrzyn adobrzyn is a code owner

@mgawarkiewicz-intel mgawarkiewicz-intel Awaiting requested review from mgawarkiewicz-intel mgawarkiewicz-intel is a code owner

@afierka-intel afierka-intel Awaiting requested review from afierka-intel afierka-intel is a code owner

@michalkuligowski michalkuligowski Awaiting requested review from michalkuligowski michalkuligowski is a code owner

@iboiko-habana iboiko-habana Awaiting requested review from iboiko-habana iboiko-habana is a code owner

@kamil-kaczor kamil-kaczor Awaiting requested review from kamil-kaczor kamil-kaczor is a code owner

@ksmusz ksmusz Awaiting requested review from ksmusz ksmusz is a code owner

Assignees

@xuechendi xuechendi

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

3 participants

@pi314ever @xuechendi