Skip to content

Accelerated Deep Learning with PyTorch at Jupyter Day Atlanta II

Notifications You must be signed in to change notification settings

hpcgarage/accelerated_dl_pytorch

Repository files navigation

accelerated_dl_pytorch

This is the repository for Accelerated Deep Learning with Pytotch tutorial and Jupyter Day Atlanta 2018 talk slides. It features full tutorial notebook, Jupyter Notebook Slides html file, and a demo with surface finish quality inspection.

Knowledge Prerequisites

This tutorial assumes familiarity with Python and Numpy.

Tutorial Prerequisites

Python3 is required to run this tutorial. You also will need some libraries from SciPy package (NumPy, Matplotlib, Pandas), Jupyter Notebook support, Seaborn for plotting, and Pytorch 0.3.0 or newer.

The simpliest way to maintain Python with all these libraries as well as many others is to install Anaconda. You can Find Pytorch installation instructions on the Pytorch page.

CUDA availability is not strictly required, but highly desirable. Life is short -- use a GPU!

How to Use

Our tutorial git has two submodules - for surface dataset, and for pretrained model for surface finish quality inspection. To download the tutorial, use

git clone --recurse-submodules [email protected]:hpcgarage/accelerated_dl_pytorch.git

If you didn't clone repository with its submodules, you can always clone submodules with this command:

git submodule update --init --recursive

To run the Jupyter Notebook Slides as at Jupyter Day Atlanta 2018 talk, you can use following command:

jupyter nbconvert tutorial_presentation.ipynb --to slides --post serve

Table of Contents

  1. Essential PyTorch Background
  2. PyTorch for Data Analytics
  3. LeNet Convolutional Neural Network (CNN) in PyTorch
  4. Application to a Manufacturing Problem

About

Accelerated Deep Learning with PyTorch at Jupyter Day Atlanta II

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published