-
Notifications
You must be signed in to change notification settings - Fork 3
/
dataset.py
358 lines (319 loc) · 14.6 KB
/
dataset.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
import os
import numpy as np
import pandas as pd
import pickle
import random
import torch
from numpy import genfromtxt
from collections import defaultdict
from sklearn.utils import shuffle
random.seed(7)
np.random.seed(7)
class Dataset:
def __init__(self, ds_name, args):
random.seed(7)
np.random.seed(7)
self.args = args
self.device = args.device
self.name = ds_name
self.dir = "datasets/" + self.name + "/"
self.ent2id, self.ent2id_typed, self.entid2typid = self.read_pickle(
'ent2id'), self.read_pickle('ent2id_typed'), self.read_pickle('entid2typid')
self.instype_spec, self.instype_transitive, self.instype_all = self.read_pickle(
'instype_spec'), self.read_pickle('instype_transitive'), self.read_pickle('instype_all')
self.rel2id = self.read_pickle('rel2id')
self.class2id, self.class2id2ent2id, self.subclassof2id = self.read_pickle(
'class2id'), self.read_pickle('class2id2ent2id'), self.read_pickle('subclassof2id')
self.r2id2dom2id, self.r2id2range2id = self.read_pickle(
'r2id2dom2id'), self.read_pickle('r2id2range2id')
self.r2id2metadom2id, self.r2id2metarange2id = self.read_pickle(
'r2id2metadom2id'), self.read_pickle('r2id2metarange2id')
self.setting = args.setting
self.sem = args.sem
if self.args.lossfunc == 'bce':
inv_r2id2dom2id = {}
for k, v in self.r2id2dom2id.items():
try:
inv_r2id2dom2id[k +
max(self.rel2id.values()) +
1] = self.r2id2range2id[k]
except BaseException:
pass
inv_r2id2range2id = {}
for k, v in self.r2id2range2id.items():
try:
inv_r2id2range2id[k +
max(self.rel2id.values()) +
1] = self.r2id2dom2id[k]
except BaseException:
pass
self.r2id2dom2id.update(inv_r2id2dom2id)
self.r2id2range2id.update(inv_r2id2range2id)
self.r2hs = self.read_pickle('heads2id_original')
if self.args.lossfunc == 'bce':
self.r2ts_origin = self.read_pickle('tails2id_original')
self.r2ts = self.r2ts_origin.copy()
for rel, tails in self.r2ts_origin.items():
self.r2ts[rel + len(self.rel2id)] = self.r2hs[rel]
else:
self.r2ts = self.read_pickle('tails2id_original')
self.data = {}
if self.args.lossfunc == 'bce':
self.data["pd_train"] = pd.read_csv(
self.dir +
"train2id_inv.txt",
sep='\t',
header=None,
names=[
'h',
'r',
't'])
self.data["pd_valid"] = pd.read_csv(
self.dir +
"valid2id_inv.txt",
sep='\t',
header=None,
names=[
'h',
'r',
't'])
self.data["pd_test"] = pd.read_csv(
self.dir +
"test2id_inv.txt",
sep='\t',
header=None,
names=[
'h',
'r',
't'])
self.data["train"] = self.data["pd_train"].to_numpy()
self.data["valid"] = self.data["pd_valid"].to_numpy()
self.data["test"] = self.data["pd_test"].to_numpy()
self.data["df"] = pd.concat(
[self.data["pd_train"], self.data["pd_valid"], self.data["pd_test"]]).to_numpy()
else:
self.data["pd_train"] = pd.read_csv(
self.dir +
"train2id.txt",
sep='\t',
header=None,
names=[
'h',
'r',
't'])
self.data["pd_valid"] = pd.read_csv(
self.dir +
"valid2id.txt",
sep='\t',
header=None,
names=[
'h',
'r',
't'])
self.data["pd_test"] = pd.read_csv(
self.dir +
"test2id.txt",
sep='\t',
header=None,
names=[
'h',
'r',
't'])
self.data["train"] = self.data["pd_train"].to_numpy()
self.data["valid"] = self.data["pd_valid"].to_numpy()
self.data["test"] = self.data["pd_test"].to_numpy()
self.data["df"] = pd.concat(
[self.data["pd_train"], self.data["pd_valid"], self.data["pd_test"]]).to_numpy()
self.neg_ratio = args.neg_ratio
self.neg_sampler = args.neg_sampler
self.all_rels = list(self.rel2id.values())
self.all_ents = list(self.ent2id.values())
self.batch_index = 0
self.id2ent = {v: k for k, v in self.ent2id.items()}
self.id2rel = {v: k for k, v in self.rel2id.items()}
self.id2class = {v: k for k, v in self.class2id.items()}
self.dumb_hr_, self.dumb_tr_ = self.read_pickle(
'dumb_hr'), self.read_pickle('dumb_tr')
self.sem_hr_, self.sem_tr_ = self.read_pickle(
'sem_hr'), self.read_pickle('sem_tr')
all_possible_hs, all_possible_ts = self.get_observed_triples(
self.data["train"], self.data["valid"], self.data["test"])
if self.args.lossfunc == 'bce':
with open(self.dir + 'pickle/observed_heads_inv.pkl', 'wb') as f:
pickle.dump(all_possible_hs, f)
with open(self.dir + 'pickle/observed_tails_inv.pkl', 'wb') as f:
pickle.dump(all_possible_ts, f)
else:
with open(self.dir + 'pickle/observed_heads_original_kg.pkl', 'wb') as f:
pickle.dump(all_possible_hs, f)
with open(self.dir + 'pickle/observed_tails_original_kg.pkl', 'wb') as f:
pickle.dump(all_possible_ts, f)
self.data["train"] = shuffle(self.data["train"], random_state=7)
def get_observed_triples(self, train2id, valid2id, test2id):
all_possible_hs = defaultdict(dict)
all_possible_ts = defaultdict(dict)
train2id = torch.as_tensor(train2id, dtype=torch.int32)
valid2id = torch.as_tensor(valid2id, dtype=torch.int32)
test2id = torch.as_tensor(test2id, dtype=torch.int32)
all_triples = torch.cat((train2id, valid2id, test2id))
X = all_triples.detach().clone()
for triple in range(X.shape[0]):
h, r, t = X[triple][0].item(
), X[triple][1].item(), X[triple][2].item()
try:
all_possible_ts[h][r].append(t)
except KeyError:
all_possible_ts[h][r] = [t]
for triple in range(X.shape[0]):
h, r, t = X[triple][0].item(
), X[triple][1].item(), X[triple][2].item()
try:
all_possible_hs[t][r].append(h)
except KeyError:
all_possible_hs[t][r] = [h]
all_possible_ts = dict(all_possible_ts)
all_possible_hs = dict(all_possible_hs)
return all_possible_hs, all_possible_ts
def lookup_chunks(self):
return len([filename for filename in os.listdir(
self.dir + "pickle/") if filename.startswith("sem_tr_" + str(self.neg_sem))])
def nb_neg(self):
sem = [
filename for filename in os.listdir(
self.dir +
"pickle/") if filename.startswith("sem_tr_")]
dumb = [
filename for filename in os.listdir(
self.dir +
"pickle/") if filename.startswith("dumb_tr_")]
return int(sem[-1][7:8])
def filtering(self):
print('train shape before filtering: ', self.data["train"].shape[0])
self.data["pd_train"] = self.data["pd_train"][(self.data["pd_train"]['r'].isin(
self.r2id2dom2id.keys())) & (self.data["pd_train"]['r'].isin(self.r2id2range2id.keys()))]
print(
'train shape after filtering 0: ',
self.data["pd_train"].shape[0])
all_rels = self.data["pd_train"]['r'].unique()
rels_suppr = []
pb_dom = list(set(self.r2id2dom2id.values()) -
set(self.class2id2ent2id.keys()))
pb_range = list(set(self.r2id2range2id.values()) -
set(self.class2id2ent2id.keys()))
for r in all_rels:
if self.r2id2dom2id[r] in pb_dom:
rels_suppr.append(r)
if self.r2id2range2id[r] in pb_range:
rels_suppr.append(r)
rels_suppr = list(set(rels_suppr))
self.data["pd_train"] = self.data["pd_train"][~self.data["pd_train"]['r'].isin(
rels_suppr)]
print(
'train shape after filtering 1: ',
self.data["pd_train"].shape[0])
if self.name != 'DBpedia77k':
self.data["pd_train"] = self.data["pd_train"][((self.data["pd_train"]['h'].isin(
self.sem_hr_.keys())) & (self.data["pd_train"]['t'].isin(self.sem_tr_.keys())))]
print(
'train shape after filtering 2: ',
self.data["pd_train"].shape[0])
self.data["train"] = self.data["pd_train"].to_numpy()
idx_pb = []
for i, triple in enumerate(self.data["train"]):
h, r, t = triple[0], triple[1], triple[2]
if self.name != 'FB15k187' and self.name != 'Yago14k':
for i in range(1, self.nb_chunks + 1):
if (h not in globals()[f'self.sem_hr_{self.neg_sem}_{i}']) or (
t not in globals()[f'self.sem_tr_{self.neg_sem}_{i}']):
idx_pb.append(i)
elif (r not in globals()[f'self.sem_hr_{self.neg_sem}_{i}'][h]) or (r not in globals()[f'self.sem_tr_{self.neg_sem}_{i}'][t]):
idx_pb.append(i)
else:
if (h not in self.sem_hr_) or (t not in self.sem_tr_):
idx_pb.append(i)
if (r not in self.sem_hr_[h]) or (r not in self.sem_tr_[t]):
idx_pb.append(i)
bad_df = self.data["pd_train"].index.isin(idx_pb)
self.data["pd_train"] = self.data["pd_train"].loc[~bad_df]
print('train shape after filtering: ', self.data["pd_train"].shape[0])
keep_ents = list(set(list(self.data["pd_train"]['h'].unique(
)) + list(self.data["pd_train"]['t'].unique())))
keep_rels = list(self.data["pd_train"]['r'].unique())
print('valid shape before filtering: ', self.data["pd_valid"].shape[0])
self.data["pd_valid"] = self.data["pd_valid"][self.data["pd_valid"]['r'].isin(
keep_rels) & self.data["pd_valid"]['h'].isin(keep_ents) & self.data["pd_valid"]['t'].isin(keep_ents)]
print('valid shape after filtering: ', self.data["pd_valid"].shape[0])
print('test shape before filtering: ', self.data["pd_test"].shape[0])
self.data["pd_test"] = self.data["pd_test"][self.data["pd_test"]['r'].isin(
keep_rels) & self.data["pd_test"]['h'].isin(keep_ents) & self.data["pd_test"]['t'].isin(keep_ents)]
print('test shape after filtering: ', self.data["pd_test"].shape[0])
print('total entities: ', len(keep_ents))
print('total relations: ', len(keep_rels))
return self.data["pd_train"].to_numpy(
), self.data["pd_valid"].to_numpy(), self.data["pd_test"].to_numpy()
def read_pickle(self, file):
try:
with open(self.dir + "pickle/" + file + ".pkl", 'rb') as f:
pckl = pickle.load(f)
return pckl
except BaseException:
print(file + ".pkl not found.")
def num_ent(self):
return len(self.ent2id)
def num_class(self):
return len(self.class2id)
def num_rel(self):
return len(self.rel2id)
def was_last_batch(self):
return (self.batch_index == 0)
def next_pos_batch(self, batch_size):
if self.batch_index + batch_size < len(self.data["train"]):
batch = self.data["train"][self.batch_index: self.batch_index + batch_size]
self.batch_index += batch_size
else:
batch = self.data["train"][self.batch_index:]
self.batch_index = 0
np.random.shuffle(batch)
return batch
def sem_picking(self, neg_batch, side):
idx_tails = np.where(side == 2)[0].tolist()
idx_heads = np.where(side == 0)[0].tolist()
for i, triple in enumerate(neg_batch[np.where(side == 2)]):
try:
neg_batch[idx_tails[i], 2] = np.random.choice(
self.sem_hr_[triple[0]][triple[1]])
except ValueError:
neg_batch[idx_tails[i], 0] = np.random.choice(
self.sem_tr_[triple[2]][triple[1]])
except KeyError:
neg_batch[idx_tails[i], 2] = np.random.choice(
self.dumb_hr_[triple[0]][triple[1]])
for i, triple in enumerate(neg_batch[np.where(side == 0)]):
try:
neg_batch[idx_heads[i], 0] = np.random.choice(
self.sem_tr_[triple[2]][triple[1]])
except ValueError:
neg_batch[idx_heads[i], 2] = np.random.choice(
self.sem_hr_[triple[0]][triple[1]])
except KeyError:
neg_batch[idx_heads[i], 0] = np.random.choice(
self.dumb_tr_[triple[2]][triple[1]])
return neg_batch
def dumb_picking(self, neg_batch, side):
idx_tails = np.where(side == 2)[0].tolist()
idx_heads = np.where(side == 0)[0].tolist()
for i, triple in enumerate(neg_batch[np.where(side == 2)]):
neg_batch[idx_tails[i], 2] = np.random.choice(
self.dumb_hr_[triple[0]][triple[1]])
for i, triple in enumerate(neg_batch[np.where(side == 0)]):
neg_batch[idx_heads[i], 0] = np.random.choice(
self.dumb_tr_[triple[2]][triple[1]])
return neg_batch
def neg_picking(self, pos_batch, neg_ratio=1):
neg_batch_sem = np.repeat(np.copy(pos_batch), neg_ratio, axis=0)
neg_batch_dumb = np.repeat(np.copy(pos_batch), neg_ratio, axis=0)
M = neg_batch_sem.shape[0]
side = np.random.choice([0, 2], size=M)
neg_batch_sem = self.sem_picking(neg_batch_sem, side)
neg_batch_dumb = self.dumb_picking(neg_batch_dumb, side)
return neg_batch_sem, neg_batch_dumb