Repository navigation
headless: matrix Jobs at default priority, print transcripts per cell (#157) - #158
Merged
Merged
Conversation
…#157) PRIORITY_CLASS for run-matrix-k8s.sh (empty = cluster default, the line is omitted). run_matrix.sh EMIT_TRANSCRIPTS=1 prints each cell's transcript under the existing ----- <name>.json ----- marker as soon as the cell finishes; the Job sets it and drops the end-of-run dump, so a preempted Job keeps every finished cell. SKIP_CELLS reruns only the missing cells.
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Closes #157.
run-matrix-k8s.shtakes a newPRIORITY_CLASSvariable. Empty (the default) omitspriorityClassName, so matrix Jobs run at the cluster default.PRIORITY_CLASS=opportunisticremains available for short, massively parallel work.run_matrix.shgainsEMIT_TRANSCRIPTS=1(set by the Job), which prints each cell's transcript under the existing----- qN__model__tK.json -----marker as soon as the cell finishes. The end-of-run dump is removed so each transcript appears exactly once (glen-benchmark'scollect_run.pydoes not deduplicate). A Job that stops early keeps every finished cell.SKIP_CELLS(space-separatedqN__model__tKnames) skips cells a stopped Job already finished.README, AGENTS.md and CHANGELOG updated.
Testing: rendered the template with and without
PRIORITY_CLASS(line dropped when empty, YAML valid; invalid value rejected). Ranrun_matrix.shwith a stubnodeand parsed the output with glen-benchmark'stranscripts_from_logandFAIL_RE: 3/3 transcripts, 2/2 FAILs, skipped cell skipped. Not yet run as a Job on the cluster.