3. Sync results back¶
Pulls the outputs from the cluster to your machine.
Safe by design: pull() fetches only output/ and done/. It never
touches notebooks/ or scripts/, so if you edited a notebook locally while
jobs were running on the HPC, syncing results will not overwrite your change.
(Requires rsync on your machine.)
import nb2slurm
wf, cfg = nb2slurm.load_config("control_config.json")
Pull the results¶
wf.pull(ssh=cfg, dry_run=True) # preview the rsync commands
wf.pull(ssh=cfg) # for real
Changed a notebook? Re-push before re-submitting¶
If you edited something in notebooks/, re-run 1_build.ipynb (its wf.push
step) to upload the new version, then submit again from 2_submit.ipynb.
Finished jobs are recorded in done/done.csv, so re-submitting only runs what’s
new or unfinished.