Deep Learning with PyTorch Quick Start Guide

Building a single-layer CNN

So now we should have enough theory to build a simple convolution network and understand how it works. Here is a model class template we can start with:

The basic convolutional unit we will be using is in PyTorch is the nn.Conv2d module. It is characterized by the following signature:

nn.Conv2d(in_channels, outs_channels, kernel_size, stride=1, padding = 0)

The values of these parameters are constrained by the size of the input data and the formulae discussed in the last section. In this example, in_channels is set to 1. This refers to the fact that our input image has one color dimension. If we were working with ...