pvphan/pytorchtutorial

PyTorch starter code for solving MNIST

PyTorch Tutorial

(Figure from Abhishek Paudel [5], linked below)

This template provides a quick way to experiment with neural network architectures on the classic 'MNIST database of handwritten digits' using PyTorch. Library dependencies and dataset parsing are handled for you.

Prerequisites on your local machine are docker, make, wget, gunzip.

Quick start:

1. Clone this repo and cd into it.
2. In mnist.py, modify the class MnistModel and hyperparameters. You can use pytorch.py as an example.
3. In console, run make runmnist to train and test the network. It will print Correctly predicted __% to tell you how the network performed.
4. Repeat from step 2!

References:

1. (dataset) MNIST database of handwritten digits
2. (videos) 3Blue1Brown: Neural Networks
3. (tutorial) Linear Regression with PyTorch
4. (tutorial) PyTorch basics
5. (blog) Abhishek Paudel: Pen Stroke Sequence Feature Extraction from MNIST Digits
6. (repo) https://github.com/anibali/docker-pytorch