Skip to content

Commit c727f15

Browse files
authored
Update 04-cgroups-and-gres.md
1 parent ab10200 commit c727f15

File tree

1 file changed

+6
-6
lines changed

1 file changed

+6
-6
lines changed

admin-training/slurm/04-cgroups-and-gres.md

Lines changed: 6 additions & 6 deletions
Original file line numberDiff line numberDiff line change
@@ -104,18 +104,18 @@ In order to enable GRES plugins, a comma delimited list of GRES must be provided
104104
* *GresTypes* defines the list of generic resources to be managed.
105105
* *Gres* defines the resource configuration details in the format <name>[:<type>][:no_consume]:<number>[K|M|G]
106106

107-
### Configure support for bandwidth
107+
### Configure support for a custom GRES (AI ASIC chips)
108108

109109
Include bandwidth GRES in GresTypes parameter in the slurm.conf file:
110110

111111
```
112-
GresTypes=bandwidth
112+
GresTypes=myriad
113113
```
114114

115-
Setup the nodes.conf file to extend the nodes definition with ```Gres=bandwidth:no_consume:4G```:
115+
Setup the nodes.conf file to extend the nodes definition with ```Gres=myriad:no_consume:4```:
116116

117117
```
118-
NodeName=hsw[001-512] NodeAddr=slurm-simulator NodeHostName=slurm-simulator RealMemory=258048 Sockets=2 CoresPerSocket=12 ThreadsPerCore=1 State=UNKNOWN Feature=hsw Gres=bandwidth:no_consume:4G
118+
NodeName=hsw[001-512] NodeAddr=slurm-simulator NodeHostName=slurm-simulator RealMemory=258048 Sockets=2 CoresPerSocket=12 ThreadsPerCore=1 State=UNKNOWN Feature=hsw Gres=myriad:no_consume:4
119119
NodeName=skl[001-512] NodeAddr=slurm-simulator NodeHostName=slurm-simulator RealMemory=389120 Sockets=2 CoresPerSocket=20 ThreadsPerCore=1 State=UNKNOWN Feature=skl
120120
NodeName=knl[0001-1024] NodeAddr=slurm-simulator NodeHostName=slurm-simulator RealMemory=389120 Sockets=1 CoresPerSocket=72 ThreadsPerCore=4 State=UNKNOWN Feature=knl
121121
```
@@ -127,7 +127,7 @@ Modify gres.conf file to define the nodes that have this special resource
127127

128128
```
129129
# Configure support for our four GPUs, plus bandwidth
130-
NodeName=hsw[001-512] Name=bandwidth Count=4G
130+
NodeName=hsw[001-512] Name=myriad Count=4
131131
```
132132

133133
Finally, restart the Slurm daemons to apply the changes.
@@ -139,5 +139,5 @@ In order to request a specific GRES, the users need to use ```--gres``` or ```#S
139139
The option requires an argument specifying which generic resources are required and how many resources. The resource specification is of the form name[:type:count].
140140

141141
```
142-
su - user01 -c "sbatch -n 512 --gres=bandwidth:2G --wrap='env; srun -n 1 sleep 120'"
142+
su - user01 -c "sbatch -n 512 --gres=myriad:2 --wrap='env; srun -n 1 sleep 120'"
143143
```

0 commit comments

Comments
 (0)