CI Pathways: Intro to PyTorch

PyTorch is widely recognized for its flexibility and ease of use, making it a top choice in both academic research and industry applications. In this session, we will establish foundational concepts of neural networks, understand their training process, and practice the concepts of automatic differentiation and backpropagation. Participants will gain an understanding of essential components of neural network training, such as loss functions, optimizers, and key architectural components. They will learn how to implement a basic neural network using PyTorch and train it on GPUs using the MNIST dataset. This training aims to build both theoretical knowledge and practical skills, equipping participants to advance in the field of deep learning.

Pre-requisites:

• Basic Python programming, such as using NumPy
• Basic linear algebra (matrix-matrix multiplication)
• To participate in the hands-on exercises, you must know the basics of using NCSA's Delta

CI Pathways is a training program led by the National Center for Supercomputing Applications and the Pittsburgh Supercomputing Center funded by NSF award 2417789. For more information about the program, please visit the CI Pathways webpage on HPC-Moodle.