run.py

N_EPOCHS = 5
best_valid_loss = float('inf')

for epoch in range(N_EPOCHS):
     
    #train the model
    train_loss, train_acc = train(model, train_iterator, optimizer, criterion)
    
    #evaluate the model
    valid_loss, valid_acc = evaluate(model, valid_iterator, criterion)
    
    #save the best model
    if valid_loss < best_valid_loss:
        best_valid_loss = valid_loss
        torch.save(model.state_dict(), 'saved_weights.pt')
    
    print(f'\tTrain Loss: {train_loss:.3f} | Train Acc: {train_acc*100:.2f}%')
    print(f'\t Val. Loss: {valid_loss:.3f} |  Val. Acc: {valid_acc*100:.2f}%')