Skip to content

Explorers Group: TF 2.X and PyTorch for not so Dummies

  • Date: Tuesday, 21 Feb 2020
  • Time: 1600
  • Location: IPL OE
  • Lead: J. Emmanuel Johnson

Blurb

TensorFlow (TF) is one of the most popular automatic differentiation (autograd) libraries in the world right now being used in production as well as research. Backed by Google (or Alphabet, Inc.), it has a huge company support system and the open-source community is massive so there is a ton of code available for a lot of the state-of-the-art (SOTA) machine learning algorithms. A more recent autograd library, PyTorch, was founded by Facebook and has risen to be the second most popular autograd library available. It's overtaken TensorFlow in the research industry because it is more Pythonic and it uses dynamic graphs which better suited the research community. TensorFlow recently got an update (TF 2.X) that has integrated keras (a high-level TF wrapper) and incorporated some more PyTorch-like design principles. So we will be going over some of the key features that you need to know to get you started on your Machine Learning and/or Deep Learning journey using TensorFlow and/or PyTorch.

What To Expect

This will be fairly high level but we will have some code examples. I will be presenting an overview of the nature/status of deep learning software and then we will be going through a google colab notebook outlining some of the key features of TensorFlow and PyTorch and answering any questions people have. It should take more more than 1.5 hours. Some other things: This will be in Python so familiarity with the language is expected and/or a strong familiarity with programming. I expect some level of machine learning and/or deep learning background to be able to keep up with some of the terminology, e.g. optimization, loss function, etc. This will not be a live coding session but you're more than welcome to bring your own laptops. Even if you think this is too high-level (or low-level) but are still interested, check the resources I've listed below. There might be some useful tutorials for you.

Resources

I have put the materials online. Please go through them if you get a chance. In particular I will be giving a short introduction about the nature of a deep learning library and we will spend the rest of the time walking through the key features of this colab notebook. The resources I've linked will be updated over the weekend and over the next few months so check back for updates if you're still interested.

Spirit Animal

Did you know that some Koala bears carry a strand of chlamydia (Herpes)? Some news reports even got so far as to kill them due to potential human infections. A moment of silence for the remaining 70% of the population of our moody, furry, cuddly friends as they deal with the crazy Austrailian wildfires.