for i in range(nclasses):
        print("SVM Number %d out of %d" % (i+1, nclasses))
        binary_training_set_y = [x if x == float(i+1) else 0.0 for x in training_set_y]
        binary_test_set_y = [x if x == float(i+1) else 0.0 for x in test_set_y]

        prob = svm_problem(binary_training_set_y, training_set_x)
        param = svm_parameter(paramlist)
        aux_name = base_name + str(i+1)

        if load == 0:
            model = svm_train(prob, param)
            svm_save_model(aux_name, model)
        else:
            svm_load_model(aux_name)

        if operation == 1:
            p_labels, p_acc, p_vals = svm_predict(binary_test_set_y, test_set_x, model)

            all_p_vals.append(np.squeeze(p_vals))
            all_p_labels.append(np.squeeze(p_labels))

if operation == 1:
        multi_class_labels = []

        for i in range(len(test_set_y)):
            local_max = -math.inf
            index_max = 0

            for j in range(nclasses):
                if all_p_vals[j][i] >= 0.0:
                    if all_p_vals[j][i] > local_max:
                        local_max = all_p_vals[j][i]
                        index_max = j+1
                else:
                    if all_p_labels[j][i] > 0.0:
                        if abs(all_p_vals[j][i]) > local_max:
                            local_max = abs(all_p_vals[j][i])
                            index_max = j+1

            multi_class_labels.append(float(index_max))