def get_accuracy(p, threshold):
    net = Net()
    net.load_state_dict(torch.load('./models/p=' + str(p) + '_t=' + str(threshold)))
    net.eval()
    total = 0
    second_total = 0
    class_correct = list(0. for i in range(10))
    class_total = list(0. for i in range(10))
    with torch.no_grad():
        for data in testloader:
            images, labels = data
            outputs = net(images, is_trained = True)
            _, predicted = torch.max(outputs, 1)
            c = (predicted == labels).squeeze()
            label = labels[0]
            class_correct[label] += c.item()
            class_total[label] += 1
    accuracy = sum(class_correct)/sum(class_total) * 100
    return accuracy