-
Notifications
You must be signed in to change notification settings - Fork 1
/
submitClusterTestJob.sh
executable file
·43 lines (36 loc) · 1.99 KB
/
submitClusterTestJob.sh
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
#!/bin/sh
# usage: submitClusterJob.sh [WORKERS]
# check if #workers were provided
if [ $# -eq 0 ]
then
WORKERS=16
elif [ $# -eq 1 ]
then
WORKERS=$1
fi
# standard cluster call
# use --mem-per-cpu=5120 for runs with traintime > 1 mio --reservation=lelasch --mem-per-cpu=3072,
# available until next friday: autumnchat 23908 --mem-per-cpu=1494 for 16 cores, fletcher 23972, springtalk 48164
# arguments for testing procedure: 'GammaVsMetCosts_0,5mio', 'CritLRVsMetCosts_1mio'
srun --partition=sleuths --reservation=lelasch --nodelist=autumnchat --cpus-per-task=$WORKERS --gres gpu:2 -LXserver \
--job-name=GammaVsMetCosts \
matlab -nodisplay -r "parTest({'GammaVsMetCosts_0,5mio'}, ${WORKERS}, 1); quit"
#srun --partition=sleuths --reservation=lelasch --nodelist=fletcher --cpus-per-task=$WORKERS --mem-per-cpu=1498 --gres gpu:2 -LXserver \
#--job-name=GammaVsMetCosts \
#matlab -nodisplay -r "parTest({'CritLRVsMetCosts_1mio'}, ${WORKERS}, 1); quit"
# srun --partition=sleuths --reservation=lelasch --nodelist=springtalk --cpus-per-task=$WORKERS --mem-per-cpu=3010 --gres gpu:2 -LXserver \
# --job-name=CritLRVsMetCosts \
# matlab -nodisplay -r "parTest({'CritLRVsMetCosts_1mio'}, ${WORKERS}, 1); quit"
# srun --partition=sleuths --nodelist=vane --cpus-per-task=$WORKERS --mem-per-cpu=5120 --gres gpu:2 -LXserver \
# --job-name=CritLRVsMetCosts \
# matlab -nodisplay -r "parTest({'CritLRVsMetCosts_1mio'}, ${WORKERS}, 1); quit"
# the time command can track the cpu usage
# srun --partition=sleuths --cpus-per-task=$WORKERS --mem-per-cpu=3072 --gres gpu:2 -LXserver \
# time matlab -nodisplay -r "parOES(1); quit"
# possible outcome of the time command:
# 191.01user 30.59system 35:04:55elapsed 0%CPU (0avgtext+0avgdata 515960maxresident)k
# 14424inputs+92808outputs (5major+1873105minor)pagefaults 0swaps
# alternative cluster call (batch job)
# SBATCH --partition=sleuths --cpus-per-task=$WORKERS --mem-per-cpu=3072 --gres gpu:2 -LXserver\
# --job-name=GammaVsMetCosts \
# matlab -nodisplay -r "parOES(${WORKERS}); quit"