Presented By O'Reilly and Cloudera
Make Data Work
March 13–14, 2017: Training
March 14–16, 2017: Tutorials & Conference
San Jose, CA

PyTorch: A flexible and intuitive framework for deep learning

James Bradbury (Salesforce Research)
2:40pm3:20pm Wednesday, March 15, 2017
Secondary topics:  AI, Deep learning
Average rating: ****.
(4.00, 8 ratings)

What you'll learn

  • Explore PyTorch, a brand-new deep learning framework intended to be faster, easier, and more flexible than alternatives like TensorFlow


The last few years have seen an explosion of interest in deep learning, but a data scientist new to the field faces an overwhelming array of open source software frameworks to choose from.

James Bradbury makes the case for PyTorch, a brand-new deep learning framework from developers at Facebook AI Research, Twitter Cortex, and Salesforce Research that’s intended to be faster, easier, and more flexible than alternatives like TensorFlow. PyTorch is based on the efficient and well-tested Torch backend, but with a Python frontend built from the ground up for intuitive, rapid prototyping of new deep learning models for image, text, and time series data.

James explains the define-by-run approach that makes PyTorch different and outlines examples from the fields of natural language processing and reinforcement learning that demonstrate its power and simplicity. Code will be made available for all examples used in the talk.

Photo of James Bradbury

James Bradbury

Salesforce Research

James Bradbury is a research scientist at Salesforce Research, where he works on cutting-edge deep learning models for natural language processing. James joined Salesforce with the April 2016 acquisition of MetaMind Inc., where he designed and implemented a neural machine translation system that won second place in the WMT 2016 machine-translation competition. He is a contributor to the Chainer and PyTorch deep learning software frameworks. James holds a degree in linguistics from Stanford University.