Advanced Deep Learning with Python

Training

We'll implement the training itself with the help of the train function, which takes the network and the cuda device as parameters. We'll use cross-entropy loss and the Adam optimizer (the usual combination for classification tasks). The function simply iterates EPOCHS times and calls the train_epoch and test functions for each epoch. The following is the implementation:

def train(model: torch.nn.Module, device: torch.device):    loss_function = torch.nn.CrossEntropyLoss()    optimizer = torch.optim.Adam(model.parameters())    train_loader, val_order = create_datasets()  # read datasets    # train    for epoch in range(EPOCHS):        print('Epoch {}/{}'.format(epoch + 1, EPOCHS))        train_epoch(model, device, loss_function, optimizer, train_loader) test(model ...