Skip to content

docs(changelog): add v2.1.2 release notes for #149, #150, #154#155

Merged
coketaste merged 1 commit into
developfrom
coketaste/changelog-v2.1.2
Jul 13, 2026
Merged

docs(changelog): add v2.1.2 release notes for #149, #150, #154#155
coketaste merged 1 commit into
developfrom
coketaste/changelog-v2.1.2

Conversation

@coketaste

Copy link
Copy Markdown
Collaborator

Summary

  • Add a v2.1.2 CHANGELOG section documenting the three fixes on develop ahead of main
  • Brings in the sglang-disagg co-located proxy fix, the MAD_SECRETS_* redaction fix, and the log error-pattern/performance-metrics priority fix

Changes covered

Test plan

Document the three fixes on develop ahead of main: log error-pattern
match no longer overrides valid performance metrics (#154), sglang-disagg
co-located proxy support (#149), and MAD_SECRETS_* redaction in printed
commands (#150).

Co-Authored-By: Claude Opus 4 <noreply@anthropic.com>
@coketaste coketaste self-assigned this Jul 13, 2026
Copilot AI review requested due to automatic review settings July 13, 2026 15:54
@coketaste
coketaste merged commit 6d49427 into develop Jul 13, 2026

Copilot AI left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

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

Pull request overview

Adds v2.1.2 release notes to the project changelog to document three fixes already landed on develop (log error-pattern vs perf-metrics priority, sglang-disagg 2-node co-located proxy topology support + default split fix, and MAD_SECRETS_* redaction in logged Docker commands).

Changes:

  • Add a new 2.1.2 section dated 2026-07-13 with entries for #149, #150, and #154.
  • Document behavior changes for log scanning, SLURM launcher topology validation/default split, and secret redaction.

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

Comment thread CHANGELOG.md

### Changed

- **Log error-pattern match no longer overrides valid performance metrics** (#154, ROCM-27774): The post-run log scanner greps the entire run log, including a model's own generated stdout, so a generative benchmark (e.g. an LLM emitting `"ValueError:"` in a code sample) could flip an otherwise-successful run to `FAILURE`. Since the scan cannot distinguish framework/harness diagnostics from model output, valid extracted performance metrics now take priority: a pattern match no longer fails a run that produced valid perf data — the match is still surfaced (in yellow) for triage instead of silently failing. Status logic was extracted into a new `resolve_run_status()` helper (`execution/container_runner_helpers.py`) with unit tests. When no valid metrics exist, a pattern match still fails the run as before.
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.

2 participants