Skip to content

misc: update cuda graph capture exception log - #894

Merged
zhyncs merged 2 commits into
sgl-project:mainfrom
zhyncs:upd
Aug 2, 2024
Merged

zhyncs merged 2 commits into
sgl-project:mainfrom
zhyncs:upd

Conversation

@zhyncs

@zhyncs zhyncs commented Aug 2, 2024

Copy link
Copy Markdown
Contributor

Thank you for your contribution, we really appreciate it. The following instructions will help improve your pull request and make it easier to receive feedback. If there are any items you don't understand, don't worry. Just submit the pull request and ask the maintainers for help.

Motivation

Please explain the motivation behind this PR and the goal you aim to achieve with it.

Modification

Briefly describe the changes made in this PR.

Checklist

  1. Ensure pre-commit pre-commit run --all-files or other linting tools are used to fix potential lint issues.
  2. Confirm that modifications are covered by complete unit tests. If not, please add more unit tests for correctness.
  3. Modify documentation as needed, such as docstrings or example tutorials.

@zhyncs

zhyncs commented Aug 2, 2024

Copy link
Copy Markdown
Contributor Author

ref #867 (comment)

@zhyncs
zhyncs merged commit 6b8f66e into sgl-project:main Aug 2, 2024
@zhyncs
zhyncs deleted the upd branch August 2, 2024 14:40
timethink pushed a commit to timethink/sglang that referenced this pull request Mar 9, 2025
efschu pushed a commit to efschu/htsglang that referenced this pull request Aug 26, 2026
…at lost in silence: end S3, S4 and S5) into the flip train
efschu pushed a commit to efschu/htsglang that referenced this pull request Aug 26, 2026
…own merges moved

The merge-regression run over the day's ticket suites turned up exactly one
failure outside the recorded standing set:
test_rejected_evidence_pins.py::PpWithSpecEvidenceTest::
test_evidence_cites_land_on_BOTH_halves_of_the_guard.

Not a defect in any merged fix. sgl-project#898 re-pinned this row against
be56c4e, and the same day's server_args.py additions -- sgl-project#889, sgl-project#893,
sgl-project#894, sgl-project#897 and sgl-project#887/sgl-project#890 -- then pushed the guard down 44 lines. The
verdict did not change; its coordinates did. :19436/:19451 -> :19480/:19495,
both re-read in the merged tree rather than taken from the failure text.

This is the guard doing its job: a line pin is a coordinate into a file
other tickets edit, so it goes stale at MERGE time, not at write time, and
only a train run can see it. #704b re-pinned the same row for the same
reason, so the note now says so instead of reading as a one-off.

CLASS. test_rejected_evidence_pins.py is the only line-pin validator in the
tree (grep for _LINE_REF / "cite has drifted"), and its other five cases
pass, so no sibling pin drifted unseen behind a missing guard. 6 passed.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant