new_scores = scores.get_scores(net, refrigerator_test_set, 1,
params.REFRIGERATOR_WINDOW_SIZE, std, mean)
if scores.compare_scores(best_scores, new_scores) > 0:
best_model.load_state_dict(net.state_dict())
best_scores = new_scores
torch.save(best_model, 'models/refrigerator_trained_model.pt')
print('Best trained model')
print('-------------------------------------------\n\n')
print('Finished Training')
torch.save(best_model, 'models/refrigerator_trained_model.pt')
net.eval()
last_scores = scores.get_scores(net, refrigerator_test_set, 1,
params.REFRIGERATOR_WINDOW_SIZE, std, mean)
dataiter = iter(refrigerator_testloader)
for i in range(len(refrigerator_testloader)):
aggregate, labels = dataiter.next()
label = labels.mean(dim=1).float()
inputs = aggregate
inputs = inputs.view(-1, 1, params.REFRIGERATOR_WINDOW_SIZE)
if torch.cuda.is_available():
inputs, labels = inputs.cuda(), labels.cuda()
inputs = Variable(inputs.float())
outputs = net(inputs)
#print('Outputs:', outputs)
for i in range(len(inputs)):
data.show_output(aggregate=aggregate[i],
individual=labels[i],
output=outputs.data[i][0],
label=label[i],
window_size=params.REFRIGERATOR_WINDOW_SIZE)
torchvision.transforms.Compose([Normalize(mean=mean, sd=std)]))
print('Size of test set: ', len(test_set))
# del training_set
net = ConvDishNILM()
try:
net = torch.load('models/dishwasher_lucky_shot_house1.pt')
except FileNotFoundError:
print('There is no pretrained model')
if torch.cuda.is_available():
net = net.cuda()
print('Testing.')
print("Trained network's results: ")
net_scores = train.test_network(net, test_set, params.DISHWASHER_WINDOW_SIZE,
std, mean)
for i in range(len(test_set)):
aggregate, labels = test_set[i]
label = labels.mean()
inputs = aggregate
inputs = inputs.view(-1, 1, params.DISHWASHER_WINDOW_SIZE)
if torch.cuda.is_available():
inputs, labels = inputs.cuda(), labels.cuda()
inputs = Variable(inputs.float())
outputs = net(inputs)
#print('Outputs:', outputs)
data.show_output(aggregate=aggregate,
individual=labels,
output=outputs.data[0][0],
label=label,
window_size=params.DISHWASHER_WINDOW_SIZE)