.

scripts/python/make_gcts.py

@@ -458,9 +458,9 @@ def generate_filtered_transects(
 
     logging.info(f"Transects written to {TMP_BASE_URI}")
 
-    logging.info("Restarting client...")
-    client.restart()
-    logging.info(f"Client dashboard link: {client.dashboard_link}")
+    # logging.info("Restarting client...")
+    # client.restart()
+    # logging.info(f"Client dashboard link: {client.dashboard_link}")
 
     transects = dask_geopandas.read_parquet(
         TMP_BASE_URI, storage_options=storage_options

src/coastpy/utils/dask.py

@@ -86,12 +86,13 @@ def _create_slurm_client(self, *args: Any, **kwargs: Any) -> Client:
         """
         from dask_jobqueue import SLURMCluster
 
-        # Define default values specific to SLURM
+
         slurm_configs = {
-            # "cores": 1,  # Cores per worker
-            # "processes": True,  # Processes per worker
+            "cores": 5,  # Cores per worker
+            "processes": 5,  # Processes per worker
             # "n_workers": 5,
-            # "memory": "12GB",  # Memory per worker
+            "memory": "60GB",  # Memory per worker
+            # job_script_prologue=["module load devel/cuda/12.1"],
             # "local_directory": "/scratch/frcalkoen/tmp",
             # "walltime": "1:00:00",
         }
@@ -100,7 +101,7 @@ def _create_slurm_client(self, *args: Any, **kwargs: Any) -> Client:
 
         # Create the SLURM cluster
         cluster = SLURMCluster(*args, **slurm_configs)
-        cluster.scale(jobs=5)
+        cluster.scale(jobs=1)
 
         logging.info(f"{cluster.job_script()}")