/pytorch-demos

PyTorch Demos used in presentations

Primary LanguageJupyter Notebook

Pytorch Demos

This is a set of PyTorch demos that I initially prepared for the Developer Circle Leads Summit in Lagos, December 2018

I expect to be able to grow this repository of demos

Prerequisits

  • Install Anaconda
  • Install pyTorch
  • Install Jupyter
  • Additional packages to install (use conda install)
    • torchvision
    • numpy
    • matplotlib

Cat Demo

This shows how to train a simple perceptron to answer the question "Can we estimate if a Facebook post contains a cat picture based solely on the amounts of likes and shares"? The dataset does not contain any real data and it's a simple artificial clustered dataset that simulates that the more people share and like a post, the more likely it is that it has a cat picutre.

Neural Network Demo

All the pyTorch complexities are hidden away in helper modules, which allows for demoing a Neural Network being trained and classifying, without getting hung up on details. As a follow up to the presentation, we can show the audience the inner workings of the helper modules.