Computing with GPUs

Requesting GPUs

GPUs are requested via the --gpus option:

sbatch --gpus=1 ...

Specific GPU models (see table below) can be requested by indicating model and count:

sbatch --gpus=rtx2080ti:1 ...

Nodes GPUs

Location	Partition	Node Name	GPU
FHCRC	all	j	NVIDIA GTX 1080ti
FHCRC	all	k	NVIDIA RTX 2080ti
FHCRC	none (interactive use)	rhino	NVIDIA RTX1080ti

Using GPUs

When your job is assigned a GPU, Slurm sets the environment variable CUDA_VISIBLE_DEVICES. This environment variable indicates the assigned GPU- most CUDA tools (e.g. tensorflow) use this to restrict execution to that device.