Advanced Deep Learning with Python

Putting it all together

Finally, we can run the whole thing. The full code for this is available at https://github.com/PacktPublishing/Advanced-Deep-Learning-with-Python/blob/master/Chapter11/imitation_learning/main.py.

The following snippet builds and restores (if available) the network, runs the training, and evaluates the network:

# create cuda devicedev = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")# create the networkmodel = build_network()# if true, try to restore the network from the data filerestore = Falseif restore:    model_path = os.path.join(DATA_DIR, MODEL_FILE)    model.load_state_dict(torch.load(model_path))# set the model to evaluation (and not training) modemodel.eval()# transfer to the gpumodel = model.to(dev) ...