torch.save() / torch.load()

torch.save(model.state_dict(), "data/model.pth")

model = NeuralNetwork()
model.load_state_dict(torch.load('data/model.pth'))
…
model.eval()