PSFC GPU Cluster

	Computer Services Calendar Safety Search PSFC

	PSFC GPU Cluster
Computer Services Network & Email Status Computer Account Request Network Connection Request Network Printer List PSFC VPN Remote Access Change PSFC Password PSFC Mailing Lists Introduction to the C-Mod Experiment Alcator Physics Research High-Energy-Density Physics Waves & Beams Fusion Technology & Engineering Useful Links	Overview The cluster consists of six nodes with four nvidia gpus each. Three nodes have four v100s cards and three nodes have four rtx6000 cards. The former are double precision and the latter are more appropriate for machine learning like workflows. Requesting Access To request access to the PSFC GPU cluster, send an email to: engaging@psfc.mit.edu with your name and PSFC username. Software Stack CentOS 8.2 Python 3 CUDA 10.2 (nvcc) cuDNN Tensorflow 2 with GPU support PyTorch Caffe OpenMPI 4.0.4 PGI CUDA Fortran compiler (nvfortran) Accessing the Cluster You need to be on the PSFC network in order to access the cluster. If you are not the network, you can connect to it using the PSFC VPN. Please follow the instructions at https://www-internal.psfc.mit.edu/computers/vpn.html. Log into the cluster using ssh to gpu.psfc.mit.edu. You can also access this node using the OnDemand web interace https://ood.psfc.mit.edu/. This is the gateway node and you launch your jobs from here. Note that gpu.psfc.mit.edu does not have nvidia cards or compilers. You have to request an interactive job to access a node for compilation. You can also currently login directly to one of the gpu nodes as well, but please only use that for quick compilations not running jobs. The v100s nodes are: gpu-v100s-01.psfc.mit.edu, gpu-v100s-03.psfc.mit.edu and gpu-v100s-05.psfc.mit.edu. The RTX6000 nodes are: gpu-rtx6000-02.psfc.mit.edu, gpu-rtx6000-04.psfc.mit.edu and gpu-rtx6000-06.psfc.mit.edu. Job Management SLURM is used for job management just as on engaging. An example job requesting one gpu card per node on 2 nodes: #!/bin/bash # #SBATCH --job-name=hello_world #SBATCH --output=hello.txt #SBATCH --ntasks-per-node=1 #SBATCH --nodes=2 #SBATCH --gres=gpu:1 #SBATCH --time=05:00 #SBATCH -p regular mpirun hostname mpirun nvidia-smi #----------------------------- Help Desk for the PSFC GPU Cluster, email: hpc@psfc.mit.edu Note: current documentation is non-existent, so please address any questions you have on usage to: engaging@psfc.mit.edu
	77 Massachusetts Avenue, NW16, Cambridge, MA 02139, psfc-info@mit.edu