Skip to content

Run the GPU jobs on kempner_interactive - #4

Open
mmshad wants to merge 5 commits into
mainfrom
kempner-interactive
Open

mmshad wants to merge 5 commits into
mainfrom
kempner-interactive

Conversation

@mmshad

@mmshad mmshad commented Nov 3, 2026 •

Copy link
Copy Markdown
Contributor

Adds a partition setting and steps to run the GPU jobs of the example on kempner_interactive.

  • Adds round.partition next to round.constraint. The loops pass both to every round job they submit.
  • Before a search starts, the loop job checks the round job scripts with sbatch --test-only. If SLURM would reject them, the loop stops before any agent step.
  • An array job no longer counts a wait behind its own running tasks against the queue limit, so the rounds of a wave can run in turns with wave.max_parallel: 1. The loop plans the time of a wave for the turns.
  • When a job reaches the queue limit, the loop log shows the SLURM reason, for example QOSMaxGRESPerUser.
  • The README has a new section, Run on kempner_interactive. AGENTS.md and the experiment files show the new settings.
  • Tested on kempner_interactive: the baseline, round 1, and a small run of each loop. 119 unit tests pass.
  • Add kempner_interactive steps to the agentic workflow examples kempner-computing-handbook#457 adds the same steps to the two example chapters. Merge this PR first.

Closes #3

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

documentation Improvements or additions to documentation enhancement New feature or request

Projects

None yet

Development

Successfully merging this pull request may close these issues.

[Feature]: Run the GPU jobs on kempner_interactive

1 participant