AMS: Difference between revisions

AMS (view source)

538 bytes removed , 2 months ago

no edit summary

cc_staff

120

edits

@@ Line 31: / Line 31: @@
 <!--T:14-->
 Graham uses the Slurm scheduler; for details about submitting jobs, see [[Running jobs]].
-<b>Because Graham's scratch file system is aged now, large AMS jobs may run slow on the cluster.
-For now, we recommend user to run AMS jobs using cpus within one node (one full or partial node) and use the local disk for SCM_TMPDIR.
-Each computer node has about 950GB local disk space. See example .sh file below </b>
 ====Example scripts for an adf job ==== <!--T:16-->
@@ Line 53: / Line 49: @@
 module unload openmpi
 module load ams/2024.102
-export SCM_TMPDIR=$SLURM_TMPDIR       # for cpus within 1 node and file size <900GB. comment out this line when you run large across multi-node job
+export SCM_TMPDIR=$SLURM_TMPDIR
 bash H2O_adf.run                    # run the input script
 }}
@@ Line 165: / Line 161: @@
 module unload openmpi
 module load ams/2024.102
-export SCM_TMPDIR=$SLURM_TMPDIR       # for cpus within 1 node and file size <900GB. comment out this line when you run large across multi-node job
+export SCM_TMPDIR=$SLURM_TMPDIR
 bash SnO_EFG_band.run                 # run the input file
 }}