Add TORCH_NIGHTLY flag for full CI run against torch nightly#382
Open
atalman wants to merge 1 commit into
Open
Add TORCH_NIGHTLY flag for full CI run against torch nightly#382atalman wants to merge 1 commit into
atalman wants to merge 1 commit into
Conversation
Today NIGHTLY=1 only runs the curated subset of steps tagged with mirror.torch_nightly against the torch nightly image. This adds a new TORCH_NIGHTLY flag (env var or the ready-torch-nightly PR label) that builds and runs the *entire* test suite against torch nightly, in addition to a full run on the pinned torch. - global_config: read TORCH_NIGHTLY env var / ready-torch-nightly label; force run_all so the pinned-torch suite runs fully too. - buildkite_step: when set, collect every command step into the 'vLLM Against PyTorch Nightly' group and auto-run it (unblocked). - bootstrap-amd/intel: plumb the flag and label through. - docs + unit test for the full-run behavior.
Collaborator
Author
|
Companion vLLM-repo PR (wires |
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Purpose
Today
NIGHTLY=1only runs a curated subset of steps (those taggedmirror: {torch_nightly: {}}) against the torch-nightly image. There is no way to run the full CI suite against torch nightly.This PR adds a new
TORCH_NIGHTLYflag that builds and runs the entire test suite against torch nightly, in addition to a full run on the pinned torch — effectively a full "vLLM vs torch nightly" signal on demand.Triggered by either:
TORCH_NIGHTLY=1, orready-torch-nightlyPR label.Behavior
When
TORCH_NIGHTLY=1:vLLM Against PyTorch Nightlygroup and auto-run (unblocked) against the torch-nightly image (image_build_torch_nightly.sh, CUDA 13,PYTORCH_NIGHTLY=1) — not just themirror.torch_nightly-tagged subset.run_allis forced so the full suite also runs on the pinned torch image.NIGHTLY=1TORCH_NIGHTLY=1Changes
pipeline_generator/global_config.py: readTORCH_NIGHTLYenv var andready-torch-nightlylabel; forcerun_all.pipeline_generator/buildkite_step.py: collect all steps into the nightly group and setauto_runwhen the flag is set.bootstrap-amd.sh/bootstrap-intel.sh: plumb the flag + label through.pipeline_generator/README.md: document the new env var.tests/pipeline_generator/test_step.py: new test asserting an untagged step runs unblocked in the nightly group underTORCH_NIGHTLY=1.Test Plan
pytest tests/pipeline_generator/— 6 passed.--torch-nightlyintotrigger-ci-build.shand documents the flag.Authored with AI assistance.