Skip to content

Support Python 3.6 in Slurm provenance writers#824

Open
jder wants to merge 1 commit into
mainfrom
codex/fix-apptainer-python36-provenance
Open

Support Python 3.6 in Slurm provenance writers#824
jder wants to merge 1 commit into
mainfrom
codex/fix-apptainer-python36-provenance

Conversation

@jder

@jder jder commented Jul 24, 2026

Copy link
Copy Markdown
Member

Summary

  • keep the Apptainer train and eval provenance writers compatible with ORCD's host Python 3.6
  • replace str.removeprefix with equivalent prefix slicing in both Slurm harnesses
  • document why the older-Python-compatible form is required

Root cause

The Slurm harness writes run-provenance.json with the host python3 before launching Apptainer. ORCD currently provides Python 3.6.8 on the host, where str.removeprefix is unavailable. As a result, H200 smoke-test job 18753331 pulled the correct container successfully but exited before srun launched training.

Impact

Training and evaluation jobs can now write provenance metadata and proceed to their container launch on ORCD's host Python.

Checks

  • bash -n scripts/slurm_apptainer_train.sbatch scripts/slurm_apptainer_eval.sbatch
  • executed each harness's embedded provenance writer with ORCD host Python 3.6.8 and verified the expected JSON keys
  • git diff --check

@jder
jder marked this pull request as ready for review July 24, 2026 18:15
@jder
jder enabled auto-merge July 24, 2026 18:15
@jder
jder requested a review from alxmrs July 24, 2026 18:15
@jder

jder commented Jul 24, 2026

Copy link
Copy Markdown
Member Author

I'm not sure why the default python on this host is one that was EOL in 2023 but this seems like an easy fix at least.

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