def blend(s1_num, ind_f, ind_t, connection):
    s2_num = 1.0 - s1_num
    blended_array = []
    blended_vec = []
    lt = len(template)

    for i in range(ind_f, ind_t):    
        p = np.array(nlp.vocab[s1[i]].vector)
        o = np.array(nlp.vocab[s2[i]].vector)


        p *= s1_num
        o *= s2_num

        blended_vec = p + o


        ms = nlp.vocab.vectors.most_similar(np.asarray([blended_vec]), n=16)
        words = [nlp.vocab.strings[w] for w in ms[0][0]]

    
    
        blended_array = blended_array + [words[15].lower()]

    
    connection.send(blended_array)



if __name__ == '__main__':
    conn1, conn2 = multiprocessing.Pipe()
    modified_array = []

    process1 = Process(target=blend, args=(0.25, 0, 50, conn2))
    process2 = Process(target=blend, args=(0.25, 50, 100, conn2))
    process3 = Process(target=blend, args=(0.25, 100, 150, conn2))
    process4 = Process(target=blend, args=(0.25, 150, 218, conn2))

    process1.start()

    modified_array = modified_array + conn1.recv()

    process2.start()

    modified_array = modified_array + conn1.recv()

    process3.start()

    modified_array = modified_array + conn1.recv()

    process4.start()

    modified_array = modified_array + conn1.recv()
    process1.join()
    process2.join()
    process3.join()
    process4.join()

    print_arrayed_text(modified_array)

I am writing a program that does some spacy on a text, and it involves processing vectors of 300 in an array of length 218. I’m splitting an array into 4 parts and am trying to do processing on each part in parallel then joining the array afterword. Any idea why its as slow as sequential?