forked from stanfordhpccenter/HTR-solver
-
Notifications
You must be signed in to change notification settings - Fork 0
/
galileo.slurm
42 lines (37 loc) · 960 Bytes
/
galileo.slurm
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
#!/bin/bash -eu
#SBATCH --job-name=prometeo
if [[ "$QUEUE" == "gll_usr_prod" ]]; then
USE_CUDA=0
CORES_PER_NODE=36
NUMA_PER_RANK=2
RAM_PER_NODE=70000
# Resources:
# 118GB RAM per node
# 2 NUMA domains per node
# 18 cores per NUMA domain
elif [[ "$QUEUE" == "gll_usr_gpuprod" ]]; then
USE_CUDA=1
CORES_PER_NODE=36
NUMA_PER_RANK=2
RAM_PER_NODE=70000
GPUS_PER_NODE=2
FB_PER_GPU=10000
# Resources:
# 118GB RAM per node
# 2 NUMA domains per node
# 18 cores per NUMA domain
# 1 nVidia K80 GPUs (seen as two K40 gpus)
else
echo "Unrecognized queue $QUEUE" >&2
exit 1
fi
cd $SLURM_SUBMIT_DIR
source "$HTR_DIR"/jobscripts/jobscript_shared.sh
srun -n "$NUM_RANKS" --ntasks-per-node="$RANKS_PER_NODE" --cpus-per-task="$CORES_PER_RANK" \
--export=ALL \
$COMMAND
# Resources:
# 118GB RAM per node
# 2 NUMA domains per node
# 18 cores per NUMA domain
# 1 nVidia K80 GPUs (seen as two K40 gpus)