Distributed Deep Learning with GPUs on HPC

Deep Learning (DL) is a powerful tool transforming scientific workflow. Because DL training is computationally intensive, it is well suited to HPC systems. Effective use of UVA HPC resources can greatly accelerate your DL workflows. In this tutorial, we will discuss:

  • When is it appropriate to use a GPU?
  • How to optimize single-GPU code?
  • How to convert single-GPU code to Multi-GPU code in different frameworks and run it on UVA HPC?

Please download and unzip the code to follow along with the activities.

distributed_dl.zip