HPC

Useful instructions / commands for the HPC shell. For starting please check the Techne Site.

Quick Start here.

Table of Contents

Accessing the HPC

There are majorly 3 ways of accessing the HPC provided at Red Hen Lab:

  1. OnDemand Portal: The OnDemand portal on the web here provides access to the HPC cluster and is the best possible way to access anything on the HPC. It not only provides a shell access, but also access to directories where files can be downloaded using a GUI, a desktop access (with/without GPU as per requirement), and some features such as jupyter notebooks, and some more.
  2. Terminal Access from local computer: It is done with the help of a VPN. Same as the HPC shell access. For the VPN setup, check here. *The format for the password for VPN access to HPC is as follows: hpc_passphrase,duo_code which needs to be followed for successful connection through VPN. Connection certificate is not required and can be ignored.
  3. X2Go GUI Interface: It is a desktop application that provides access to the HPC. Check here for more info.

1 and 2 have worked for me. I haven’t had success with 3. However, for all the 3 ways, Duo authentication (similar to Google Authenticator) is required. Follow all the steps except the VPN step here to get the Duo setup.

Useful Shell Commands

Most of the commands are similar to any Linux bash commands. I will list a few important/required during the work.

Note: Root access is denied in the HPC

Available Modules

$ module avail

Access to GPU

$ module load cuda                          # Default Cuda version loaded
$ module load cuda/10.1                     # Specific Cuda version loaded
$ srun -p gpu --gpus 1 --mem 4000 --cpus-per-gpu 2 --pty bash

Access to Tensorflow

$ module load gcc
$ module load python/3.7.0
$ module load cuda/10.1
$ module load tensorrt/6
$ python
>>> import tensorflow as tf
>>> tf.__version__
2.1.0                                       # As per June 2021