Welcome to USP HPC’s documentation!

High-Performance Computing (HPC) at Universidade de Sao Paulo (USP) comprises two clusters of servers for parallel and distributed computing for scientific research. Use Python libraries to train machine and deep learning models in a single server on GPU and scale up to the full cluster for large datasets.

Common attributes:

  • Workload manager (Slurm) that schedules jobs in all servers.

  • Shared network file system visible to all servers.

  • Limited local storage.

  • Login node with no GPU in lince cluster.

Cluster aguia contains servers for CPU processing while lince cluster is for GPU processing. Access to the clusters through shark.

Both clusters allow parallel and distributed computing for scientific purposes, for example training of machine learning and deep learning models.

This documentation explains how to:

  1. Setup a development environment for testing and debugging.

  2. Schedule and manage Slurm jobs.

  3. Use Python libraries to train machine learning and deep learning models.

This is a quick start guide for new users and may save several hours of searching and testing. Detailed and complete information on each topic is available in the Internet.