Pytorch Tutorial

March 8th, 9:00am - 1:00pm

Annenberg 213

PyTorch talk at Caltech (Please RSVP)

Abstract:

Join Facebook's AI team to learn how PyTorch is being used to accelerate the path from novel research to large scale production deployment in computer vision, natural language processing, and machine translation. In the tutorial portion, you'll learn the basics of implementing deep learning algorithms in PyTorch. Concretely, you will walk away with a basic understanding of implementing convolutional networks, using standard gradient based optimization methods, manipulating PyTorch tensors (hint very similar to NumPy.. :)), using automatic differentiation via autograd, and how to use deep-learning specific modules.

Speakers: Joe Spisak, Francisco Massa, Brad Heintz, Shen Li, Christian Puhrsch

Agenda:

General PyTorch introduction (9:00 - 9:30).
Introductory tutorials
- Tensor (9:30 -10:00)
- Autograd (10:00 - 10:30)
- Neural Networks (10:30 - 11:00)
- Training on Cifar10 (11:00 - 11:30)
Break (11:30 - 12:00)
Self-paced tutorials
- Transfer learning (12:00 - 12:30)
- Visual understanding of CNNs (12:30 - 1:00)

Audience:

Students/Researchers on non-CS areas without prior exposure to PyTorch, but who want to get into Deep Learning.

Compute needed:

For this tutorial, we will be using Colab which requires no credit card and provides a free GPU for up to 12hrs of usage. Colab natively supports PyTorch as well as libraries like torchvision thanks to our collaboration with Google.