Requesting and using GPUs
As part of the gpu partition there are a number of GPU equipped nodes available.
Currently there are 7 nodes each with 2 NVIDIA A100 GPUs. One additional node is in the interactive partition
In order to access the GPUs they need to be requested via SLURM as one does for other resources such as CPUs and memory.
The flag required is
--gres=gpu:1 for 1 GPU per node and
--gres=gpu:2 for 2 GPUs per node.
An example job script is as follows:
#!/bin/bash #SBATCH --nodes 1 #SBATCH --ntasks 1 #SBATCH --cpus-per-task 12 #SBATCH --mem 64G #SBATCH --time 12:00:00 # NOTE - GPUS are in the gpu partition #SBATCH --partition gpu #SBATCH --gres gpu:1 #SBATCH --gres-flags enforce-binding # Set up my modules module purge module load my list of modules module load cuda # Check that the GPU is visible nvidia-smi # Run my GPU enable python code python mygpucode.py
#SBATCH --gres gpu:1 is omitted then no GPUs will be visible even if they are present on the compute node.
If you request one GPU it will always be seen as device 0.
#SBATCH --gres-flags enforce-binding option ensures that the CPUs allocated will be on the same PCI bus as the GPU(s) which greatly improves the memory bandwidth. This may mean that you have to wait longer for resources to be allocated but it is strongly recommended.
If you select 2 GPUs then we strongly advise also requesting
#SBATCH --exculsive to have all the resources of the node available to your job.
In order to use the CUDA toolkit there is a module available
module load cuda
This loads the nvcc compiler and CUDA libraries. There is also a cudnn nodule for the DNN tools/libraries
Containers and GPUs
Singularity containers can make use of GPUs but in order to make them visible to the container environment an extra flag "--nv" must be passed to Singularity
module load singularity singularity run --nv mycontainer.sif
The full documentation is at https://sylabs.io/guides/3.5/user-guide/gpu.html