Example for HPC cluster @ ITU

This is a PyTorch implementation of a deep learning approach for digit recognition (classification) with MNIST dataset.

1. Setting up a virtual environment with Anaconda

For loading "conda":

$ module load Anaconda3/2021.05

Then, for adding the configurations to your .bashrc file.

$ conda init bash

Then, you can build a virtual environment to install your packages as follows.

$ conda create -n nn_env pytorch torchvision torchaudio cudatoolkit=11.3 -c pytorch

2. Cloning the repository

$ git clone https://github.com/ameliajimenez/example-mnist-itu.git
$ cd example-mnist-itu/

Notice that you have to change the working directory with --workdir argument in example_slurm.sh file.

3. Launching the script

$ sbatch example_slurm.sh

4. Monitoring

We can monitor the running tasks on the cluster by the following command:

$ squeue

or in a continuous way:

$ watch squeue

Documentation

ITU's HPC Documentation Project