The generic consumable resources are :
#SBATCH --gres=gpu:1
Keep in mind for gpu's you need to load the module of the required cuda version !
Once you request a gpu(s) resource, the scheduler will set for you the environment variable : CUDA_VISIBLE_DEVICES
This will point to the assigned gpu(s) for your job, you shall only use those and not others !!!
Some gpu boards have the nvlink modules fitted, this will allow you to double the gpu memory and cumputing power. If you request two gpu's with nvlink, you need to force socket binding using the following option :
#SBATCH --sockets-per-node=1
The available features are :
For example to force only a40 gpu's
#SBATCH --constraint=a40
The HPC/SLURM cluster contains multiple common partitions :
| Partition name | Nodes | Details | available to |
| main | ctit[080-094],caserta,hpc-node[01-07] | All | |
| debug | ctit[080-094],caserta,hpc-node[01-12,14-19] | admin |
As well as multiple additional partitions :
| Partition name | Nodes | Details | available to |
| am | hpc-node[01-04] | eemcs-(dmmp/macs/mast/mia/mms/sor/stat) | |
| bdsi | ctit087 | bms-bdsi | |
| bmpi | hpc-node[15-16] | gpu | tnw-bmpi |
| bss | hpc-node15 | cpu | eemcs-bss |
| dmb | ctit[084-085,092],hpc-node[07,09] | eemcs-dmb | |
| mia | ctit[090-091,093-094],hpc-node05 | eemcs-mia | |
| mia-pof | hpc-node06 | eemcs-mia & tnw-pof | |
| tfe | hpc-node[16-19] | cpu | et-tfe |
| ps | hpc-node[11-12,14] | eemcs-ps | |
| ram | ctit[086,089] | eemcs-ram | |
| students | hpc-node08 | eemcs-students |
The main partition is the default partition that will be used to submit a job to any of the nodes. The debug partition is for testing purposes only.
Access to the additional are limited to the funders during the first year of investment, these can be reached using the funders partitions.
Including multiple partitions is also possible. For example :
#SBATCH --partition=main,dmb
#SBATCH --partition=main,am,mia
#SBATCH --partition=main,students
See the EEMCS-HPC Hardware page for all the partition definitions.