 Now we want you to design the training loop. In tutorial one you constructed your own network and it's great that it worked. Now it is time to construct the other part of the deep learning code, namely your training loop. What goes into a training loop? The first thing is you need to choose the correct criteria, the loss function that you want to minimize. Then you need to code up the training part. You need to calculate gradients for the loss and you need to step forward. And then you need to keep track of the running loss for each epoch. We want to know the average loss of the batch size and we've already done the same thing for accuracy for you. So now it's your turn. Build your own training loop. Now do not rush, take your time. This is truly important.