write() argument must be str, not bytes　を解決したい

興味のある論文の公開されているソースコードを実行しようとしたところwrite() argument must be str, not bytes のエラーに悩まされており，解決できそうにないため，お力を貸していただきたいです．

あまり，新しい論文ではないため，python2系で書かれていたコードをpython3で書き換えている途中です．

ぜひ，よろしくお願いします．

以下がそのコードです。

発生している問題・エラーメッセージ

TypeError                                 Traceback (most recent call last)
<ipython-input-20-4999cc12e722> in <module>
     78 f = open('../data/FB15k_entity2idx.pkl', 'w')
     79 g = open('../data/FB15k_idx2entity.pkl', 'w')
---> 80 cPickle.dump(entity2idx, f, -1)
     81 cPickle.dump(idx2entity, g, -1)
     82 f.close()

TypeError: write() argument must be str, not bytes

該当のソースコード

python
1import os, sys
2#import cPickle
3import _pickle as cPickle
4
5import numpy as np
6import scipy.sparse as sp
7
8# Put the freebase15k data absolute path here
9datapath = "D:/Downloads/dataset/"
10assert datapath is not None
11
12if 'data' not in os.listdir('../'):
13    os.mkdir('../data')
14
15
16def parseline(line):
17    lhs, rel, rhs = line.split('\t')
18    lhs = lhs.split(' ')
19    rhs = rhs.split(' ')
20    rel = rel.split(' ')
21    return lhs, rel, rhs
22
23#################################################
24### Creation of the entities/indices dictionnaries
25
26np.random.seed(753)
27
28entleftlist = []
29entrightlist = []
30rellist = []
31
32for datatyp in ['train']:
33    f = open(datapath + 'freebase_mtr100_mte100-%s.txt' % datatyp, 'r')
34    dat = f.readlines()
35    f.close()
36    for i in dat:
37        lhs, rel, rhs = parseline(i[:-1])
38        entleftlist += [lhs[0]]
39        entrightlist += [rhs[0]]
40        rellist += [rel[0]]
41
42entleftset = np.sort(list(set(entleftlist) - set(entrightlist)))
43entsharedset = np.sort(list(set(entleftlist) & set(entrightlist)))
44entrightset = np.sort(list(set(entrightlist) - set(entleftlist)))
45relset = np.sort(list(set(rellist)))
46
47entity2idx = {}
48idx2entity = {}
49
50
51# we keep the entities specific to one side of the triplets contiguous
52idx = 0
53for i in entrightset:
54    entity2idx[i] = idx
55    idx2entity[idx] = i
56    idx += 1
57nbright = idx
58for i in entsharedset:
59    entity2idx[i] = idx
60    idx2entity[idx] = i
61    idx += 1
62nbshared = idx - nbright
63for i in entleftset:
64    entity2idx[i] = idx
65    idx2entity[idx] = i
66    idx += 1
67nbleft = idx - (nbshared + nbright)
68
69print ("# of only_left/shared/only_right entities: ", nbleft, '/', nbshared, '/', nbright)
70# add relations at the end of the dictionary
71for i in relset:
72    entity2idx[i] = idx
73    idx2entity[idx] = i
74    idx += 1
75nbrel = idx - (nbright + nbshared + nbleft)
76print ("Number of relations: ", nbrel)
77
78f = open('../data/FB15k_entity2idx.pkl', 'w')
79g = open('../data/FB15k_idx2entity.pkl', 'w')
80cPickle.dump(entity2idx, f, -1)
81cPickle.dump(idx2entity, g, -1)
82f.close()
83g.close()
84
85#################################################
86### Creation of the dataset files
87
88unseen_ents=[]
89remove_tst_ex=[]
90
91for datatyp in ['train', 'valid', 'test']:
92    print (datatyp)
93    f = open(datapath + 'freebase_mtr100_mte100-%s.txt' % datatyp, 'r')
94    dat = f.readlines()
95    f.close()
96
97    # Declare the dataset variables
98    inpl = sp.lil_matrix((np.max(entity2idx.values()) + 1, len(dat)),
99            dtype='float32')
100    inpr = sp.lil_matrix((np.max(entity2idx.values()) + 1, len(dat)),
101            dtype='float32')
102    inpo = sp.lil_matrix((np.max(entity2idx.values()) + 1, len(dat)),
103            dtype='float32')
104    # Fill the sparse matrices
105    ct = 0
106    for i in dat:
107        lhs, rel, rhs = parseline(i[:-1])
108        if lhs[0] in entity2idx and rhs[0] in entity2idx and rel[0] in entity2idx: 
109            inpl[entity2idx[lhs[0]], ct] = 1
110            inpr[entity2idx[rhs[0]], ct] = 1
111            inpo[entity2idx[rel[0]], ct] = 1
112            ct += 1
113        else:
114            if lhs[0] in entity2idx:
115                unseen_ents+=[lhs[0]]
116            if rel[0] in entity2idx:
117                unseen_ents+=[rel[0]]
118            if rhs[0] in entity2idx:
119                unseen_ents+=[rhs[0]]
120            remove_tst_ex+=[i[:-1]]
121
122    # Save the datasets
123    if 'data' not in os.listdir('../'):
124        os.mkdir('../data')
125    f = open('../data/FB15k-%s-lhs.pkl' % datatyp, 'w')
126    g = open('../data/FB15k-%s-rhs.pkl' % datatyp, 'w')
127    h = open('../data/FB15k-%s-rel.pkl' % datatyp, 'w')
128    cPickle.dump(inpl.tocsr(), f, -1)
129    cPickle.dump(inpr.tocsr(), g, -1)
130    cPickle.dump(inpo.tocsr(), h, -1)
131    f.close()
132    g.close()
133    h.close()
134
135unseen_ents=list(set(unseen_ents))
136print (len(unseen_ents))
137remove_tst_ex=list(set(remove_tst_ex))
138print (len(remove_tst_ex))
139
140for i in remove_tst_ex:
141    print (i)

行動規範の内容に同意します

回答1件

patch
1-f = open('../data/FB15k_entity2idx.pkl', 'w')
2-g = open('../data/FB15k_idx2entity.pkl', 'w')
3+f = open('../data/FB15k_entity2idx.pkl', 'wb')
4+g = open('../data/FB15k_idx2entity.pkl', 'wb')

あと、cPickleは古いです。python3では素直にpickleを使ってください。

投稿2019/07/25 11:39

hayataka2049

総合スコア30933

jyohonoob

2019/07/26 04:55 編集

回答ありがとうございます． fとgの前についている-や+はどのような意味があるのでしょうか？また，hayataka2049さんの回答をそのまま，以前の 78 f = open('../data/FB15k_entity2idx.pkl', 'w') 79 g = open('../data/FB15k_idx2entity.pkl', 'w') この文を -f = open('../data/FB15k_entity2idx.pkl', 'w') -g = open('../data/FB15k_idx2entity.pkl', 'w') +f = open('../data/FB15k_entity2idx.pkl', 'wb') +g = open('../data/FB15k_idx2entity.pkl', 'wb') この文に置き換えて実行してみたところ，エラー文として， File "<ipython-input-9-2acda53d9cef>", line 78 -f = open('../data/FB15k_entity2idx.pkl', 'w') ^ SyntaxError: can't assign to operator が出てきました．

hayataka2049

2019/07/26 06:18

-の行をけして同じ位置に+の行を足す

jyohonoob

2019/07/26 07:05

本当に何度も，何度もすいません．また，エラーが出てしまいました． File "FB15k_parse.py", line 84 +f = open('../data/FB15k_entity2idx.pkl', 'wb') ^ SyntaxError: can't assign to operator よろしくお願いします．