Deep Learning with PyTorch

Vectorization

Data comes in various formats such as text, sound, images, and video. The very first thing that needs to be done is to convert the data into PyTorch tensors. In the previous example, we used torchvision utility functions to convert Python Imaging Library (PIL) images into a Tensor object, though most of the complexity is abstracted away by the PyTorch torchvision libraries. In Chapter 7, Generative Networks, when we deal with recurrent neural networks (RNNs), we will see how text data can be converted into PyTorch tensors. For problems involving structured data, the data is already present in a vectorized format; all we need to do is convert them into PyTorch tensors.