Python K Medoids

def kmedoids(X, k,
           starting_medoids=None,
           max_steps=50):
    if starting_medoids is None:
        medoids = init_medoids(X, k)
        print(medoids)
    else:
        medoids = starting_medoids
       
    converged = False
    labels = np.zeros(len(X))
    i = 1
   
    # Delete a file if it exists
    if os.path.exists("KMedoids_Results.txt"):
        os.remove("KMedoids_Results.txt")   
   
    f = open("KMedoids_Results.txt", "a")
    start_time = time.time()
    f.write('The program starting time ',start_time)
    while (not converged) and (i <= max_steps):
        old_medoids = np.copy(medoids)
        # print('The old medoid value is ',old_medoids)
        S = compute_distance_1(X, medoids)
        # print(S)
        labels = assign_cluster_labels(S)
        # print(labels)
         medoids = update_cluster_medoids_intercost(X,medoids,S,labels,2)
        # print('updated medoids',medoids)
        converged = has_converged(old_medoids, medoids)
        #print(converged)
        print ("iteration", i, "Centers are  = ",medoids)
        i += 1
    f.write('The program completion time',time.time())
   

    return labels,medoids

def update_cluster_medoids_intercost(X, medoids,S,labels,p):
   
    r,c=X.shape
    out_medoids = np.copy(medoids)
    total_cost=np.sum(np.min(S,axis=1))
   
    for i in set(labels):
       
        cluster_points = np.unique(X[labels == i],axis=0)
       
        for datap in cluster_points:
            old_medoid=medoids[i]
            # compute the new cost with swapping the medoid with new point
            medoids[i]=datap
            # Compute the cost with the new medoid whether it improves the distance or not
            new_cost= np.sum(np.min(compute_distance_1(X,medoids),axis=1))
            #print('The new cost is ',new_cost)
            #print('The new medoid is ',datap)                           
            if new_cost < total_cost :
                #accept the data point swap for medoid
                total_cost = new_cost
                out_medoids[i] = datap
                #print('The medoids are',medoids)
               
            else:
                # reverse the datapoint swap
                medoids[i]=old_medoid
                #print('Inside else part medoids',medoids)
     
    return out_medoids

def update_cluster_medoids_intracost(X, medoids,labels,k,p):
    
    r, c = X.shape
    medoids=np.zeros((k,c))
    # print('the k value is ',k)
    for i in range(0,k):
        medoid_idx=labels==i
        cluster=X[medoid_idx]
        if(cluster.size)==0:
            break
        total_cost=np.unique(cluster,axis=0)
        total_cost=sc.pdist(total_cost,'euclidean')
        total_cost=sc.squareform(total_cost)
        total_cost=total_cost.sum(axis=1)
        minid=np.argmin(total_cost,axis=0)
        medoids[i]=cluster[minid]
    return medoids

testing_points=np.array([[7,6],
            [2,6],
            [3,8],
            [8,5],
            [7,4],
            [4,7],
            [6,2],
            [7,3],
            [6,4],
            [3,4]
           ])
starting_medoids=np.array([[4,7],[8,5]])