Support for %j in slurm log retrieval (#252)

hemildesai · web-flow · commit 75bc3f571411 · 2025-06-02T11:03:38.000-07:00
Signed-off-by: Hemil Desai &lt;hemild@nvidia.com&gt;
diff --git a/nemo_run/run/torchx_backend/schedulers/slurm.py b/nemo_run/run/torchx_backend/schedulers/slurm.py
@@ -360,6 +360,7 @@ def _check_finished(self) -> None:
                         if self._ls_term
                         else os.path.join(self._remote_dir, f"log*{extension}")
                     )
+                    ls_term = ls_term.replace("%j", self._app_id)
                     ls_output = self._scheduler.tunnel.run(
                         f"ls -1 {ls_term} 2> /dev/null",
                         warn=True,

Original file line number	Diff line number	Diff line change
`@@ -360,6 +360,7 @@ def _check_finished(self) -> None:`
`360`	`360`	`if self._ls_term`
`361`	`361`	`else os.path.join(self._remote_dir, f"log*{extension}")`
`362`	`362`	`)`
	`363`	`+ ls_term = ls_term.replace("%j", self._app_id)`
`363`	`364`	`ls_output = self._scheduler.tunnel.run(`
`364`	`365`	`f"ls -1 {ls_term} 2> /dev/null",`
`365`	`366`	`warn=True,`