3. Define the model We can create a model in PyTorch in multiple ways - we will be creating models by subclassing nn.Module, as shown in this tutorial B. You can use any of the layers defined in the torch.nn module . In previous lessons, we have covered linear, activation, convolutional, and pooling layers, so those might be a good starting point. While constructing your neural network, remember to make sure that the dimensions of the output of a layer and the expected input of the next layer match. The final two layers of your neural network should transform every sample into a vector of 10 elements (1 for each class in the input dataset), and perform a log softmax of the vector. These layers have already been provided for you in the code given below. You can use the following code as a starting point for your neural network: import torch.nn as nn class YourModel (nn.Module): definit__(self): super()._init_() # define your layers here self.final_layer = nn. Linear (INPUT_SIZE_GOES HERE, 10) def forward(self,x): # your code output return output = nn.functional.log_softmax(self.final_layer(x), dim=1) Once again, you are free to create any model you think would work well - however, remember that creating extremely large models will cause training to take a long time, making some of the questions in section 8 hard to answer. Ideally, your network should contain less than 250,000 parameters in total. You can find out how many parameters a model has using the following lines of code: model = YourModel() print (sum( [torch.prod (torch.tensor(i.shape)) for i in model.parameters()]))