Fix array job handling in Slurm

lxm3/xm_cluster/execution/slurm.py

@@ -51,7 +51,6 @@ def _create_job_header(
  job_log_dir: str,
  job_name: str,
  ) -> str:
- num_array_tasks = None
  job_header = header_from_executor(
  job_name, executor, num_array_tasks, job_log_dir
  )
@@ -129,9 +128,10 @@ def launch(self, job_name: str, job: job_script_builder.ClusterJob):
  num_jobs = len(job.env_vars)
  else:
  num_jobs = 1
-
+ console.log(f"Launching {num_jobs} job on {self._settings.hostname}")
  job_id = self._cluster.launch(job_script_path)
  console.log(f"Successfully launched job {job_id}")
+
  if num_jobs > 1:
  job_ids = [f"{job_id}_{i}" for i in range(num_jobs)]
  else:
@@ -215,9 +215,9 @@ def header_from_executor(
 
  log_directory = executor.log_directory or job_log_dir
  if num_array_tasks is not None:
- stdout = os.path.join(log_directory, "slurm-%A_%a.out")
+ stdout = os.path.join(log_directory, "%x-%A_%a.out")
  else:
- stdout = os.path.join(log_directory, "slurm-%j.out")
+ stdout = os.path.join(log_directory, "%x-%j.out")
 
  header.append(f"#SBATCH --output={stdout}")