[CI][AMD] Support build-scoped ROCm images - #447
Open
AndreasKaratzas wants to merge 8 commits into
Open
Conversation
AndreasKaratzas
marked this pull request as ready for review
August 3, 2026 03:47
Signed-off-by: Andreas Karatzas <akaratza@amd.com>
AndreasKaratzas
force-pushed
the
akaratza_rocm_image_promotion_infra
branch
from
August 3, 2026 04:02
7ee41ef to
e1506ae
Compare
Signed-off-by: Andreas Karatzas <Andreas.Karatzas@amd.com>
Signed-off-by: Andreas Karatzas <Andreas.Karatzas@amd.com>
Resolve ROCm pipeline conflicts while preserving current build-scoped behavior and the promotion-capable handoff. Assisted-by: OpenAI Codex Signed-off-by: Andreas Karatzas <Andreas.Karatzas@amd.com>
Raise the native AMD tmpfs /dev/shm limit from 16 GiB to 32 GiB in both pipeline generators. Assisted-by: OpenAI Codex Signed-off-by: Andreas Karatzas <Andreas.Karatzas@amd.com>
AndreasKaratzas
force-pushed
the
akaratza_rocm_image_promotion_infra
branch
from
August 17, 2026 06:44
896ec73 to
e51373b
Compare
Signed-off-by: Andreas Karatzas <akaratza@amd.com>
AndreasKaratzas
force-pushed
the
akaratza_rocm_image_promotion_infra
branch
from
August 17, 2026 08:02
e51373b to
32d2108
Compare
…romotion_infra Signed-off-by: Andreas Karatzas <Andreas.Karatzas@amd.com> # Conflicts: # buildkite/bootstrap-amd.sh # buildkite/pipeline_generator/amd.py # buildkite/tests/pipeline_generator/test_amd.py
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
The vLLM pipeline needs a main-only, globally serialized promotion job, but the current V2 generator does not preserve its condition and concurrency fields through every rendering path. The existing timeout logic also assumes a Python-only base-refresh job is always short, while the new selector can turn any confirmed registry miss into a full rebuild. This adds validated
if_condition,concurrency, andconcurrency_groupfields and carries them through normal commands, direct AMD jobs, and AMD mirrors.if_conditionis emitted as Buildkite'siffield. The promotion job is always selected, with its condition excluding PR builds when the pipeline is uploaded.Every non-skipped base-selection job now gets the 540-minute timeout. Native and DinD AMD jobs use
rocm/vllm-dev:ci_base-build-$BUILDKITE_BUILD_ID, which is published by the companion vLLM build. This keeps each evaluation on theci_baseit selected instead of a stable or cross-build tag.