 When coding the transformer encoder, why are we extending the class nn.module? Torch modules is a class that we use when creating neural networks. It provides mechanisms to access and initialize parameters. It allows us to also toggle the utility of GPUs for training and CPUs for inference. It provides a method for the forward pass through the network. It helps us manage memory and manage gradients during training. And so when coding out neural networks, we will extend this class and override the forward method.