Introduction to the DL training process

- [Instructor] There are different ways to build your deep learning model. You can achieve it using supervised learning, unsupervised learning, or semi-supervised learning. Either way, you decide you're going to use the same pipeline to train, test, and deploy your deep learning model. The process begins with the data preparation stage. As it's name suggests, we load the generic data, which can be in many different formats, such as text, images, videos, audio files, et cetera, from an external source, and we convert it to numeric values suitable for model training. These numeric values are in form of tensors. Then tensors need to be pre-processed during transforms, and we group them with batches that can be passed into the model. The second stage is the model development stage that consists of three parts, designing the model, training the model using training data, and testing its performance. We take the data set and split it into three data sets, training data, validation data, and…