Loading, transforming and preparing data for building a deep learning model can be messy and difficult to maintain. PyTorch provides tools to ease this effort and decouples it from the training portion of your machine learning pipeline. Below is a summary of the three PyTorch tools we will highlight in this demo:

• Dataset: stores data and their corresponding labels

• Transforms: performs data manipulation to make data suitable for training

• DataLoaders: iterable around the dataset for ease of access to samples from the dataset.

Let’s dive into each of these components and load our design safe dataset.