-
Notifications
You must be signed in to change notification settings - Fork 2
/
Copy pathrun.py
1615 lines (1278 loc) · 61.2 KB
/
run.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
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
import json
import numpy as np
from helper import *
from data_loader import *
import random
from torch.utils.data import TensorDataset
# sys.path.append('./')
from model.models import *
class Runner(object):
def load_events(self):
event_type_dir = "./data/" + self.p.dataset + "/event_types.json"
event_ids_dir = "./data/" + self.p.dataset + "/event_ids.json"
event_entity_dir = "./data/" + self.p.dataset + "/event2ent.json"
self.role2id = {}
self.event_edges = {} # {event_idx: [[ent_idx1, role_idx1], [ent_idx2, role_idx2], ...]}
with open(event_type_dir, "r", encoding="utf-8") as f:
self.event_types = json.loads(f.readline())
with open(event_ids_dir, "r", encoding="utf-8") as f:
self.evt2id = json.loads(f.readline())
with open(event_entity_dir, "r", encoding="utf-8") as f:
evt2ent = json.loads(f.readline())
self.evt2ent = evt2ent
# print(evt2ent)
# read event2ent into role ids
role_num = 0
for event_id in evt2ent:
args = evt2ent[event_id]["args"]
for arg in args:
role_type = args[arg]
if role_type not in self.role2id:
self.role2id.update({role_type: role_num})
role_num += 1
for event_id in evt2ent:
args = evt2ent[event_id]["args"]
event_idx = self.evt2id[event_id]
for arg in args:
role_type_idx = self.role2id[args[arg]]
ent_idx = self.ent2id[arg]
if event_idx not in self.event_edges:
self.event_edges.update({event_idx: [[ent_idx, role_type_idx]]})
else:
self.event_edges[event_idx].append([ent_idx, role_type_idx].copy())
role_num = len(self.role2id)
self.p.role_num = role_num
self.p.event_num = len(self.evt2id)
max_role_num = max([len(evt2ent[key]["args"]) for key in evt2ent])
self.p.max_role_num = max_role_num
# print(self.event_edges)
evt2ent_list = [[] for _ in range(self.p.event_num)]
for evt_id in self.event_edges:
for ent_id in self.event_edges[evt_id]:
evt2ent_list[evt_id].append(ent_id[0])
self.evt2ent_list = evt2ent_list
# print(evt2ent_list)
# print(evt2ent_list)
# print(self.event_edges)
# print(self.role2id)
def load_data(self):
"""
Reading in raw triples and converts it into a standard format.
Parameters
----------
self.p.dataset: Takes in the name of the dataset (FB15k-237)
Returns
-------
self.ent2id: Entity to unique identifier mapping
self.id2rel: Inverse mapping of self.ent2id
self.rel2id: Relation to unique identifier mapping
self.num_ent: Number of entities in the Knowledge graph
self.num_rel: Number of relations in the Knowledge graph
self.embed_dim: Embedding dimension used
self.data['train']: Stores the triples corresponding to training dataset
self.data['valid']: Stores the triples corresponding to validation dataset
self.data['test']: Stores the triples corresponding to test dataset
self.data_iter: The dataloader for different data splits
"""
ent_set, rel_set = OrderedSet(), OrderedSet()
for split in ['train', 'test', 'valid']:
for line in open('./data/{}/{}.txt'.format(self.p.dataset, split)):
sub, rel, obj = line.strip().split('\t')
ent_set.add(sub)
rel_set.add(rel)
ent_set.add(obj)
self.p.entity_embed_dir = "./data/" + self.p.dataset + "/entity_embed.npy"
# print(self.p.entity_embed_dir)
entity_ids_dir = "./data/" + self.p.dataset + "/entity_ids.json"
entity_types_dir = "./data/" + self.p.dataset + "/entity_types.json"
relation_ids_dir = "./data/" + self.p.dataset + "/relation_ids.json"
with open(entity_ids_dir, "r", encoding="utf-8") as f:
self.ent2id = json.loads(f.readline())
with open(entity_types_dir, "r", encoding="utf-8") as f:
entity_types = json.loads(f.readline())
self.p.num_ent = len(self.ent2id)
self.p.ent_num = len(self.ent2id)
# load entity types to type_id:
self.ent_type_to_id = {}
entity_type_num = 0
for ent_id in entity_types:
if entity_types[ent_id] not in self.ent_type_to_id:
self.ent_type_to_id.update({entity_types[ent_id]: entity_type_num})
entity_type_num += 1
self.id_to_ent_type = {idx: t for t, idx in self.ent_type_to_id.items()}
self.p.entity_type_num = len(self.ent_type_to_id)
train_ent_labels, test_ent_labels, valid_ent_labels = [], [], []
train_ent_ids, test_ent_ids, valid_ent_ids = [], [], []
train_num = int(self.p.ent_num * 0.8)
valid_num = int(self.p.ent_num * 0.1) + 1
test_num = self.p.ent_num - train_num - valid_num
# print(train_num)
for ent_id in self.ent2id:
ent_idx = self.ent2id[ent_id]
if ent_idx >=0 and ent_idx < train_num:
train_ent_ids.append(ent_idx)
train_ent_labels.append(self.ent_type_to_id[entity_types[ent_id]])
elif ent_idx >= train_num and ent_idx < train_num + valid_num:
valid_ent_ids.append(ent_idx)
valid_ent_labels.append(self.ent_type_to_id[entity_types[ent_id]])
else:
test_ent_ids.append(ent_idx)
test_ent_labels.append(self.ent_type_to_id[entity_types[ent_id]])
self.train_ent_labels = torch.LongTensor(train_ent_labels).to(self.device)
self.train_ent_ids = torch.LongTensor(train_ent_ids).to(self.device)
self.valid_ent_labels = torch.LongTensor(valid_ent_labels).to(self.device)
self.valid_ent_ids = torch.LongTensor(valid_ent_ids).to(self.device)
self.test_ent_labels = torch.LongTensor(test_ent_labels).to(self.device)
self.test_ent_ids = torch.LongTensor(test_ent_ids).to(self.device)
with open(relation_ids_dir, "r", encoding="utf-8") as f:
self.rel2id = json.loads(f.readline())
# self.ent2id = {ent: idx for idx, ent in enumerate(ent_set)}
# self.rel2id = {rel: idx for idx, rel in enumerate(rel_set)}
rel2id_dict = self.rel2id.copy()
for key in rel2id_dict:
self.rel2id.update({key+'_reverse': rel2id_dict[key]+len(rel2id_dict)})
# print(self.rel2id)
self.id2ent = {idx: ent for ent, idx in self.ent2id.items()}
self.id2rel = {idx: rel for rel, idx in self.rel2id.items()}
'''
ent2id: {entity_id: 0}
rel2id: {rel_id: 0, rel_id_reverse: 1}
'''
# also load relation classification dataset
train_rel_labels, train_rel_start, train_rel_end = [], [], []
valid_rel_labels, valid_rel_start, valid_rel_end = [], [], []
test_rel_labels, test_rel_start, test_rel_end = [], [], []
self.p.num_rel = len(self.rel2id) // 2
for line in open('./data/{}/train.txt'.format(self.p.dataset)):
sub, rel, obj = line.strip().split('\t')
start_idx = self.ent2id[sub]
train_rel_start.append(start_idx)
end_idx = self.ent2id[obj]
train_rel_end.append(end_idx)
rel_idx = self.rel2id[rel]
train_rel_labels.append(rel_idx)
self.train_rel_labels = torch.LongTensor([train_rel_labels]).to(self.device).t()
self.train_rel_start = torch.LongTensor([train_rel_start]).to(self.device).t()
self.train_rel_end = torch.LongTensor([train_rel_end]).to(self.device).t()
for line in open('./data/{}/valid.txt'.format(self.p.dataset)):
sub, rel, obj = line.strip().split('\t')
start_idx = self.ent2id[sub]
valid_rel_start.append(start_idx)
end_idx = self.ent2id[obj]
valid_rel_end.append(end_idx)
rel_idx = self.rel2id[rel]
valid_rel_labels.append(rel_idx)
self.valid_rel_labels = torch.LongTensor([valid_rel_labels]).to(self.device).t()
self.valid_rel_start = torch.LongTensor([valid_rel_start]).to(self.device).t()
self.valid_rel_end = torch.LongTensor([valid_rel_end]).to(self.device).t()
for line in open('./data/{}/test.txt'.format(self.p.dataset)):
sub, rel, obj = line.strip().split('\t')
start_idx = self.ent2id[sub]
test_rel_start.append(start_idx)
end_idx = self.ent2id[obj]
test_rel_end.append(end_idx)
rel_idx = self.rel2id[rel]
test_rel_labels.append(rel_idx)
self.test_rel_labels = torch.LongTensor([test_rel_labels]).to(self.device).t()
self.test_rel_start = torch.LongTensor([test_rel_start]).to(self.device).t()
self.test_rel_end = torch.LongTensor([test_rel_end]).to(self.device).t()
# 这里的num_rel的含义是relation type的数量 而不是relation edge的数量，其实edge的数量就等于triples的数量
# num_rel在这里指的是：没有添加reverse边的时候的数量
self.p.embed_dim = self.p.k_w * self.p.k_h if self.p.embed_dim is None else self.p.embed_dim
self.data = ddict(list)
sr2o = ddict(set)
for split in ['train', 'test', 'valid']:
for line in open('./data/{}/{}.txt'.format(self.p.dataset, split)):
sub, rel, obj = line.strip().split('\t')
sub, rel, obj = self.ent2id[sub], self.rel2id[rel], self.ent2id[obj]
self.data[split].append((sub, rel, obj))
if split == 'train':
sr2o[(sub, rel)].add(obj)
sr2o[(obj, rel+self.p.num_rel)].add(sub)
# sr2o的格式：dict, {(h, r): Set({t1, t2, ...})}
# for key in sr2o:
# if len(sr2o[key]) > 1:
# print(key)
# print(sr2o[key])
# print('\n')
# print(sr2o)
self.data = dict(self.data)
# print(self.data)
# self.data: {"train": [(h, r, t), (h, r, t)], "valid": [(h, r, t), (h, r, t)]}
self.sr2o = {k: list(v) for k, v in sr2o.items()} # only contains training sr2os
# self.sr2o的格式：dict, {(h, r): [t1, t2, ...])}
for split in ['test', 'valid']:
for sub, rel, obj in self.data[split]:
sr2o[(sub, rel)].add(obj)
sr2o[(obj, rel+self.p.num_rel)].add(sub)
self.sr2o_all = {k: list(v) for k, v in sr2o.items()} # contains training, validataion, and testing sr2os
# self.sr2o_all的格式：dict, {(h, r): [t1, t2, ...])}
# print(type(self.sr2o))
# print(type(self.sr2o[(6593,9)]))
self.triples = ddict(list)
for (sub, rel), obj in self.sr2o.items():
self.triples['train'].append({'triple':(sub, rel, -1), 'obj': obj, 'label': self.sr2o[(sub, rel)], 'sub_samp': 1})
# self.triples['train']: {'triple': (h, r, -1), 'label': [t1, t2, t3,...], 'sub_samp': 1}
for split in ['test', 'valid']:
for sub, rel, obj in self.data[split]:
rel_inv = rel + self.p.num_rel
self.triples['{}_{}'.format(split, 'tail')].append({'triple': (sub, rel, obj), 'label': self.sr2o_all[(sub, rel)]})
self.triples['{}_{}'.format(split, 'head')].append({'triple': (obj, rel_inv, sub), 'label': self.sr2o_all[(obj, rel_inv)]})
# 注意：self.triples["train"] 里面只包含了训练集的所有triples，"triple"的最后一个值是-1，但是"valid", "test"里面的label包含了所有的可能的labels，(包括训练集和测试集)
self.triples = dict(self.triples)
def get_data_loader(dataset_class, split, batch_size, shuffle=True):
return DataLoader(
dataset_class(self.triples[split], self.p),
batch_size = batch_size,
shuffle = shuffle,
num_workers = max(0, self.p.num_workers),
collate_fn = dataset_class.collate_fn
)
self.data_iter = {
'train': get_data_loader(TrainDataset, 'train', self.p.batch_size),
'valid_head': get_data_loader(TestDataset, 'valid_head', self.p.batch_size, False),
'valid_tail': get_data_loader(TestDataset, 'valid_tail', self.p.batch_size, False),
'test_head': get_data_loader(TestDataset, 'test_head', self.p.batch_size, False),
'test_tail': get_data_loader(TestDataset, 'test_tail', self.p.batch_size, False),
}
self.load_events()
self.id2evt = {idx: evt for evt, idx in self.evt2id.items()}
self.edge_index, self.edge_type = self.construct_adj()
# print(self.edge_type.shape)
# print(self.edge_index.shape)
self.event_edge_index, self.event_index, self.role_type, self.role_mask, self.entity_event_index, self.entity_mask = self.construct_event_adj()
# print(self.entity_event_index.shape)
def construct_adj(self):
"""
Constructor of the runner class
Parameters
----------
Returns
-------
Constructs the adjacency matrix for GCN
"""
# edge_index: [] -- a list of tuples: [(subj1, obj1), (subj2, obj2), ...]
# edge_type: [] -- a list of edge_type_idxs: [r1,r2,...]
edge_index, edge_type = [], []
# self.data[split]: list of (h,r,t)
for sub, rel, obj in self.data['train']:
edge_index.append((sub, obj))
edge_type.append(rel)
# Adding inverse edges
for sub, rel, obj in self.data['train']:
edge_index.append((obj, sub))
edge_type.append(rel + self.p.num_rel)
# print(self.device)
edge_list = [[] for _ in range(self.p.ent_num)]
for start, end in edge_index:
edge_list[end].append(start)
# print(edge_index)
self.edge_list = edge_list
# print(edge_list)
# edge_index = torch.LongTensor(edge_index).to(self.device).t()
edge_index = torch.LongTensor(edge_index).to(self.device).t()
edge_type = torch.LongTensor(edge_type). to(self.device)
# edge_index: torch.LongTensor(2, 2*edge_num)
# edge_type: torch.LongTensor(2*edge_num)
# print(edge_type.shape)
# print(edge_index.shape)
# construct a tensor for all 1-hop previous neighbors
return edge_index, edge_type
def construct_event_adj(self):
event_index = torch.zeros(self.p.event_num, self.p.max_role_num, dtype=torch.long).to(self.device)
role_type = torch.zeros(self.p.event_num, self.p.max_role_num, dtype=torch.long).to(self.device)
role_mask = torch.zeros(self.p.event_num, self.p.max_role_num).to(self.device)
# check the maximum number of events
ent2evt = {}
for evt_idx in self.event_edges:
args = self.event_edges[evt_idx]
for arg in args:
if arg[0] not in ent2evt:
ent2evt.update({arg[0]: [evt_idx]})
else:
ent2evt[arg[0]].append(evt_idx)
self.ent2evt = ent2evt
# print(self.ent2evt)
# 改写ent2evt成为list形式
ent2evt_list = [[] for _ in range(self.p.ent_num)]
for ent_id in ent2evt:
for evt_id in ent2evt[ent_id]:
ent2evt_list[ent_id].append(evt_id)
self.ent2evt_list = ent2evt_list
# print(ent2evt_list)
# max_evt_num = max([len(ent2evt[key]) for key in ent2evt])
# entity_event_index = torch.zeros(max_evt_num, self.p.ent_num, dtype=torch.long).to(self.device)
# role_mask = torch.zeros(max_evt_num, self.p.ent_num).to(self.device)
entity_event_index = [[] for _ in range(self.p.ent_num)]
entity_mask = [[] for _ in range(self.p.ent_num)]
for event_idx in self.event_edges:
end_ents = self.event_edges[event_idx]
for i,arg in enumerate(end_ents):
event_index[event_idx][i] = arg[0]
role_type[event_idx][i] = arg[1]
role_mask[event_idx][i] = 1.0
entity_event_index[arg[0]].append(event_idx)
entity_mask[arg[0]].append(1.0)
event_index = event_index[:, 0:self.p.entity_sample_num]
role_type = role_type[:, 0:self.p.entity_sample_num]
role_mask = role_mask[:, 0:self.p.entity_sample_num]
# embed to tensor
entity_event_index_new = entity_event_index.copy()
entity_mask_new = entity_mask.copy()
max_event_num = max([len(i) for i in entity_event_index])
for i,l in enumerate(entity_event_index_new):
entity_event_index[i].extend([0] * (max_event_num - len(l)))
for i,l in enumerate(entity_mask_new):
entity_mask[i].extend([0.0] * (max_event_num - len(l)))
entity_event_index = torch.LongTensor(entity_event_index)[:, 0:self.p.event_sample_num].to(self.device)
entity_mask = torch.Tensor(entity_mask)[:, 0:self.p.event_sample_num].to(self.device)
with open("./data/" + self.p.dataset + "/temp.json", "r", encoding="utf-8") as f:
event_edge_index = json.loads(f.read())
self.event_edge_list = event_edge_index
# print(event_edge_index)
event_neigh_list = [[] for _ in range(self.p.event_num)]
for i in range(len(event_edge_index[0])):
event_neigh_list[event_edge_index[1][i]].append(event_edge_index[0][i])
self.event_neigh_list = event_neigh_list
event_edge_index = torch.LongTensor(event_edge_index).to(self.device)
return event_edge_index, event_index, role_type, role_mask, entity_event_index, entity_mask
def __init__(self, params):
"""
Constructor of the runner class
Parameters
----------
params: List of hyper-parameters of the model
Returns
-------
Creates computational graph and optimizer
"""
self.p = params
self.logger = get_logger(self.p.name, self.p.log_dir, self.p.config_dir)
self.logger.info(vars(self.p))
# pprint(vars(self.p))
# if self.p.gpu != '-1' and torch.cuda.is_available():
# self.device = torch.device('cuda:'+self.p.gpu)
# torch.cuda.set_rng_state(torch.cuda.get_rng_state())
# torch.backends.cudnn.deterministic = True
# else:
# self.device = torch.device('cpu')
if self.p.gpu != '-1' and torch.cuda.is_available():
self.device = int(self.p.gpu)
torch.cuda.set_rng_state(torch.cuda.get_rng_state())
torch.backends.cudnn.deterministic = True
else:
self.device = torch.device('cpu')
self.load_data()
self.model = self.add_model(self.p.model, self.p.score_func)
self.optimizer = self.add_optimizer(self.model.parameters())
def add_model(self, model, score_func):
"""
Creates the computational graph
Parameters
----------
model_name: Contains the model name to be created
Returns
-------
Creates the computational graph for model and initializes it
"""
model_name = '{}_{}'.format(model, score_func)
if model_name.lower() == 'compgcn_transe': model = CompGCN_TransE(self.event_edge_index, self.edge_index, self.edge_type, self.event_index, self.role_type, self.role_mask, self.entity_event_index, self.entity_mask, params=self.p)
elif model_name.lower() == 'compgcn_distmult': model = CompGCN_DistMult(self.event_edge_index, self.edge_index, self.edge_type, self.event_index, self.role_type, self.role_mask, self.entity_event_index, self.entity_mask, params=self.p)
elif model_name.lower() == 'compgcn_conve': model = CompGCN_ConvE(self.event_edge_index, self.edge_index, self.edge_type, self.event_index, self.role_type, self.role_mask, self.entity_event_index, self.entity_mask, params=self.p)
else: raise NotImplementedError
model.to(self.device)
return model
def add_optimizer(self, parameters):
"""
Creates an optimizer for training the parameters
Parameters
----------
parameters: The parameters of the model
Returns
-------
Returns an optimizer for learning the parameters of the model
"""
return torch.optim.Adam(parameters, lr=self.p.lr, weight_decay=self.p.l2)
def read_batch(self, batch, split):
# print(batch)
"""
Function to read a batch of data and move the tensors in batch to CPU/GPU
Parameters
----------
batch: the batch to process
split: (string) If split == 'train', 'valid' or 'test' split
Returns
-------
Head, Relation, Tails, labels
"""
if split == 'train':
triple, label = [ _.to(self.device) for _ in batch[0:2]]
obj_list = batch[-1]
return triple[:, 0], triple[:, 1], triple[:, 2], label, obj_list
else:
triple, label = [ _.to(self.device) for _ in batch[0:2]]
obj_list = batch[-1]
return triple[:, 0], triple[:, 1], triple[:, 2], label
def save_model(self, save_path):
"""
Function to save a model. It saves the model parameters, best validation scores,
best epoch corresponding to best validation, state of the optimizer and all arguments for the run.
Parameters
----------
save_path: path where the model is saved
Returns
-------
"""
state = {
'state_dict' : self.model.state_dict(),
# 'best_val' : self.best_val,
# 'best_epoch' : self.best_epoch,
'optimizer' : self.optimizer.state_dict(),
# 'args' : vars(self.p)
}
torch.save(state, save_path)
def load_model(self, load_path):
"""
Function to load a saved model
Parameters
----------
load_path: path to the saved model
Returns
-------
"""
state = torch.load(load_path)
state_dict = state['state_dict']
# self.best_val = state['best_val']
# self.best_val_mrr = self.best_val['mrr']
self.model.load_state_dict(state_dict)
self.optimizer.load_state_dict(state['optimizer'])
def evaluate(self, split, epoch):
"""
Function to evaluate the model on validation or test set
Parameters
----------
split: (string) If split == 'valid' then evaluate on the validation set, else the test set
epoch: (int) Current epoch count
Returns
-------
resutls: The evaluation results containing the following:
results['mr']: Average of ranks_left and ranks_right
results['mrr']: Mean Reciprocal Rank
results['hits@k']: Probability of getting the correct preodiction in top-k ranks based on predicted score
"""
if self.p.eval == "all":
left_results = self.predict_all(split=split, mode='tail_batch')
right_results = self.predict_all(split=split, mode='head_batch')
else:
left_results = self.predict_selected(split=split, mode='tail_batch')
right_results = self.predict_selected(split=split, mode='head_batch')
results = get_combined_results(left_results, right_results)
self.logger.info('[Epoch {} {}]: MRR: Tail : {:.5}, Head : {:.5}, Avg : {:.5}'.format(epoch, split, results['left_mrr'], results['right_mrr'], results['mrr']))
self.logger.info(json.dumps({key:results[key] for key in ["mrr", "mr", "hits@1", "hits@3", "hits@5", "hits@10", "hits@20"]}))
return results
def predict_all(self, split='valid', mode='tail_batch'):
"""
Function to run model evaluation for a given mode
Parameters
----------
split: (string) If split == 'valid' then evaluate on the validation set, else the test set
mode: (string): Can be 'head_batch' or 'tail_batch'
Returns
-------
resutls: The evaluation results containing the following:
results['mr']: Average of ranks_left and ranks_right
results['mrr']: Mean Reciprocal Rank
results['hits@k']: Probability of getting the correct preodiction in top-k ranks based on predicted score
"""
self.model.eval()
with torch.no_grad():
results = {}
train_iter = iter(self.data_iter['{}_{}'.format(split, mode.split('_')[0])])
total_sub, total_rel, total_obj, total_ranks = [], [], [], []
for step, batch in enumerate(train_iter):
sub, rel, obj, label = self.read_batch(batch, split)
total_sub = total_sub + sub.tolist()
total_rel = total_rel + rel.tolist()
total_obj = total_obj + obj.tolist()
# sub_index_list, init_ent_list, final_ent_list, ent_neighbors, rel_evt_idxs, evt_neighbors, new_event_index, new_entity_event_index, new_entity_mask, new_entity_list, new_entity_type = self.sample_subgraph(sub, objs)
# sub, rel, obj: torch.LongTensor(batch_size)
# label: torch.FloatTensor(batch_size, entity_num)
pred = self.model.predict(sub, rel)
# print(pred.shape)
# pred shape: (batch_size, entity_num)
b_range = torch.arange(pred.size()[0], device=self.device)
# b_range: torch.LongTensor (batch_size)
target_pred = pred[b_range, obj]
# target_pred 选取出来的正确位置的预测概率（注意是obj对应的唯一正确概率，而不是label对应的多个正确概率）
pred = torch.where(label.byte(), -torch.ones_like(pred) * 10000000, pred)
# 把label对应为1位置的pred的值换成1
pred[b_range, obj] = target_pred
# print(label.shape)
# print(torch.sum(label, 1))torch.argsort(aa, dim=1, descending=False)
ranks = 1 + torch.argsort(torch.argsort(pred, dim=1, descending=True), dim=1, descending=False)[b_range, obj]
total_ranks += ranks.tolist()
# 这个rank的意思是：因为label可以有多个，但是obj只有一个，这个rank代表的意思是，出去label中的其他项的干扰之后，obj的排名情况
ranks = ranks.float()
results['count'] = torch.numel(ranks) + results.get('count', 0.0)
results['mr'] = torch.sum(ranks).item() + results.get('mr', 0.0)
results['mrr'] = torch.sum(1.0/ranks).item() + results.get('mrr', 0.0)
for k in range(10):
results['hits@{}'.format(k+1)] = torch.numel(ranks[ranks <= (k+1)]) + results.get('hits@{}'.format(k+1), 0.0)
# print(results)
if step % 100 == 0:
self.logger.info('[{}, {} Step {}]\t{}'.format(split.title(), mode.title(), step, self.p.name))
results['ranks'] = total_ranks
results['sub'] = total_sub
results['rel'] = total_rel
results['obj'] = total_obj
return results
def generate_neg_samples(self, input_list, num):
if len(input_list) < num:
if_enough = False
new_list = input_list.copy()
full = False
while not full:
i = random.randint(0, self.p.num_ent-1)
if i not in new_list:
new_list.append(i)
if len(new_list) == num:
full = True
return new_list, if_enough
else:
if_enough = True
return input_list[0: num].copy(), if_enough
def predict_selected(self, split='valid', mode='tail_batch'):
"""
Function to run model evaluation for a given mode
Parameters
----------
split: (string) If split == 'valid' then evaluate on the validation set, else the test set
mode: (string): Can be 'head_batch' or 'tail_batch'
Returns
-------
resutls: The evaluation results containing the following:
results['mr']: Average of ranks_left and ranks_right
results['mrr']: Mean Reciprocal Rank
results['hits@k']: Probability of getting the correct preodiction in top-k ranks based on predicted score
"""
# self.model.to("cpu")
# self.model.eval()
# self.old_device = self.model.device
# # print(old_device)
# self.model.device = "cpu"
# self.model.event_conv1.device = "cpu"
# self.model.conv1.device = "cpu"
# self.device = "cpu"
# self.model.event_edge_index = self.model.event_edge_index.to("cpu")
# self.model.edge_index = self.model.edge_index.to("cpu")
# self.model.edge_type = self.model.edge_type.to("cpu")
# self.model.role_type = self.model.role_type.to("cpu")
# self.model.role_mask = self.model.role_mask.to("cpu")
# self.model.entity_event_index = self.model.entity_event_index.to("cpu")
# self.model.entity_mask = self.model.entity_mask.to("cpu")
# print("wryyyyyyyy")
with torch.no_grad():
results = {}
train_iter = iter(self.data_iter['{}_{}'.format(split, mode.split('_')[0])])
total_sub, total_rel, total_obj, total_ranks = [], [], [], []
for step, batch in enumerate(train_iter):
if self.p.dataset == "kairos":
if step > 98:
break
sub, rel, obj, label = self.read_batch(batch, split)
sub_list = sub.tolist()
obj_list = obj.tolist()
init_ent_list = []
sub_index_list = []
obj_index_list = []
node_pos_dict = {}
# rel_evt_idxs ------ final_ent_list
for j,idx in enumerate(sub_list):
if idx not in node_pos_dict:
init_ent_list.append(idx)
node_pos_dict.update({idx: len(init_ent_list) - 1})
sub_index_list.append(len(init_ent_list) - 1)
else:
sub_index_list.append(node_pos_dict[idx])
# print(sub_index_list)
# print(objs)
for j,idx in enumerate(obj_list):
if idx not in node_pos_dict:
init_ent_list.append(idx)
node_pos_dict.update({idx: len(init_ent_list) - 1})
obj_index_list.append(len(init_ent_list) - 1)
else:
obj_index_list.append(node_pos_dict[idx])
final_ent_list, ent_neighbors, rel_evt_idxs, evt_neighbors, new_event_index, new_entity_event_index, new_entity_mask, new_entity_list, new_entity_type = self.sample_subgraph(init_ent_list)
# print(len(final_ent_list))
# print(len(ent_neighbors))
# print(len(rel_evt_idxs))
# print(len(rel_evt_idxs))
selected_list, if_enough = self.generate_neg_samples(final_ent_list, 500)
if not if_enough:
final_ent_list, ent_neighbors, rel_evt_idxs, evt_neighbors, new_event_index, new_entity_event_index, new_entity_mask, new_entity_list, new_entity_type = self.sample_subgraph(selected_list)
new_label = label[:, selected_list]
# no matter how, we selected top 500
total_sub = total_sub + sub.tolist()
total_rel = total_rel + rel.tolist()
total_obj = total_obj + obj.tolist()
# sub_index_list, init_ent_list, final_ent_list, ent_neighbors, rel_evt_idxs, evt_neighbors, new_event_index, new_entity_event_index, new_entity_mask, new_entity_list, new_entity_type = self.sample_subgraph(sub, objs)
# sub, rel, obj: torch.LongTensor(batch_size)
# label: torch.FloatTensor(batch_size, entity_num)
pred = self.model.forward(sub, rel, sub_index_list, init_ent_list, final_ent_list, ent_neighbors, rel_evt_idxs, evt_neighbors, new_event_index, new_entity_event_index, new_entity_mask, new_entity_list, new_entity_type)
pred = pred[:, 0:len(selected_list)]
# print(pred.shape)
# pred shape: (batch_size, entity_num)
b_range = torch.arange(pred.size()[0], device=self.device)
# b_range = torch.arange(pred.size()[0], device="cpu")
# b_range: torch.LongTensor (batch_size)
target_pred = pred[b_range, obj_index_list]
# target_pred 选取出来的正确位置的预测概率（注意是obj对应的唯一正确概率，而不是label对应的多个正确概率）
pred = torch.where(new_label.byte(), -torch.ones_like(pred) * 10000000, pred)
# 把label对应为1位置的pred的值换成1
pred[b_range, obj_index_list] = target_pred
# print(label.shape)
# print(torch.sum(label, 1))torch.argsort(aa, dim=1, descending=False)
ranks = 1 + torch.argsort(torch.argsort(pred, dim=1, descending=True), dim=1, descending=False)[b_range, obj_index_list]
total_ranks += ranks.tolist()
# 这个rank的意思是：因为label可以有多个，但是obj只有一个，这个rank代表的意思是，出去label中的其他项的干扰之后，obj的排名情况
ranks = ranks.float()
results['count'] = torch.numel(ranks) + results.get('count', 0.0)
results['mr'] = torch.sum(ranks).item() + results.get('mr', 0.0)
results['mrr'] = torch.sum(1.0/ranks).item() + results.get('mrr', 0.0)
for k in range(10):
results['hits@{}'.format(k+1)] = torch.numel(ranks[ranks <= (k+1)]) + results.get('hits@{}'.format(k+1), 0.0)
results['hits@{}'.format(20)] = torch.numel(ranks[ranks <= (20)]) + results.get('hits@{}'.format(20), 0.0)
# print(results)
if step % 100 == 0:
self.logger.info('[{}, {} Step {}]\t{}'.format(split.title(), mode.title(), step, self.p.name))
results['ranks'] = total_ranks
results['sub'] = total_sub
results['rel'] = total_rel
results['obj'] = total_obj
# print(results)
# self.model.device = self.old_device
# self.device = self.old_device
# self.model.event_conv1.device = self.old_device
# self.model.conv1.device = self.old_device
# self.model.to(self.old_device)
# self.model.event_edge_index = self.model.event_edge_index.to(self.old_device)
# self.model.edge_index = self.model.edge_index.to(self.old_device)
# self.model.edge_type = self.model.edge_type.to(self.old_device)
# self.model.role_type = self.model.role_type.to(self.old_device)
# self.model.role_mask = self.model.role_mask.to(self.old_device)
# self.model.entity_event_index = self.model.entity_event_index.to(self.old_device)
# self.model.entity_mask = self.model.entity_mask.to(self.old_device)
return results
def sample_subgraph(self, init_ent_list):
# print(1)
rel_evt_idxs = []
for ent_idx in init_ent_list:
rel_evt_idxs = rel_evt_idxs + self.ent2evt_list[ent_idx]
rel_evt_idxs = list(set(rel_evt_idxs))
if self.p.dataset == "kairos":
rel_evt_idxs = rel_evt_idxs[0:5000]
event_ent_idxs = []
for evt_idx in rel_evt_idxs:
event_ent_idxs += self.evt2ent_list[evt_idx]
final_ent_list = init_ent_list.copy()
if self.p.dataset == "kairos":
pass
else:
for idx in event_ent_idxs:
if idx not in final_ent_list:
final_ent_list.append(idx)
ent_neighbors, evt_neighbors = final_ent_list.copy(), rel_evt_idxs.copy()
if self.p.dataset == "kairos":
for ent_idx in final_ent_list:
idxs = self.edge_list[ent_idx]
for idx in idxs:
if idx not in ent_neighbors:
ent_neighbors.append(idx)
ent_neighbors = ent_neighbors[0:10000]
else:
# ent_neighbors, evt_neighbors
for ent_idx in final_ent_list:
idxs = self.edge_list[ent_idx]
for idx in idxs:
if idx not in ent_neighbors:
ent_neighbors.append(idx)
for evt_idx in rel_evt_idxs:
idxs = self.event_neigh_list[evt_idx]
for idx in idxs:
if idx not in evt_neighbors:
evt_neighbors.append(idx)
for ent_id in ent_neighbors:
idxs = self.ent2evt_list[ent_id]
for idx in idxs:
if idx not in evt_neighbors:
evt_neighbors.append(idx)
for evt_id in evt_neighbors:
idxs = self.evt2ent_list[evt_id]
for idx in idxs:
if idx not in ent_neighbors:
ent_neighbors.append(idx)
# assert (rel_evt_idxs == evt_neighbors[0:len(rel_evt_idxs)])
# assert (final_ent_list == ent_neighbors[0:len(final_ent_list)])
# print(4)
evt_total_to_selected = [-1 for _ in range(self.p.event_num)]
for j,evt_id in enumerate(evt_neighbors):
evt_total_to_selected[evt_id] = j
# print(5)
entity_mask_list = self.entity_mask[ent_neighbors].tolist()
entity_event_index_list = self.entity_event_index[ent_neighbors].tolist()
new_entity_event_index_list = []
for i in range(len(entity_event_index_list)):
events_i = []
for j in range(len(entity_event_index_list[0])):
evt_mask = entity_mask_list[i][j]
evt_idx = entity_event_index_list[i][j]
if evt_mask != 0.0:
if evt_total_to_selected[evt_idx] == -1:
events_i.append(0)
entity_mask_list[i][j] = 0.0
else:
events_i.append(evt_total_to_selected[evt_idx])
else:
events_i.append(0)
new_entity_event_index_list.append(events_i)
# print(6)
new_entity_event_index = torch.LongTensor(new_entity_event_index_list).to(self.device)
new_entity_mask = torch.FloatTensor(entity_mask_list).to(self.device)
if self.p.dataset == "kairos":
new_event_list = [[], []]
for i in range(len(self.event_edge_list[0])):
start = self.event_edge_list[0][i]
end = self.event_edge_list[1][i]
if end in rel_evt_idxs and start in rel_evt_idxs:
new_event_list[0].append(evt_total_to_selected[start])
new_event_list[1].append(evt_total_to_selected[end])
else:
new_event_list = [[], []]
for i in range(len(self.event_edge_list[0])):
start = self.event_edge_list[0][i]
end = self.event_edge_list[1][i]
if end in rel_evt_idxs:
new_event_list[0].append(evt_total_to_selected[start])
new_event_list[1].append(evt_total_to_selected[end])
# print(7)
new_event_index = torch.LongTensor(new_event_list).to(self.device)
# calculate the entities
ent_total_to_selected = [-1 for _ in range(self.p.ent_num)]
for j,ent_id in enumerate(ent_neighbors):
ent_total_to_selected[ent_id] = j
# print(8)
# calculate the new entity edges
new_entity_list = [[], []]
new_entity_type = []
edge_index_list = self.edge_index.tolist()
edge_type_list = self.edge_type.tolist()
if self.p.dataset == "kairos":
for i in range(len(edge_index_list[0])):
start = edge_index_list[0][i]
end = edge_index_list[1][i]
if end in final_ent_list and start in ent_neighbors:
new_entity_list[0].append(ent_total_to_selected[start])
new_entity_list[1].append(ent_total_to_selected[end])
new_entity_type.append(edge_type_list[i])
else:
for i in range(len(edge_index_list[0])):
start = edge_index_list[0][i]
end = edge_index_list[1][i]
if end in final_ent_list:
new_entity_list[0].append(ent_total_to_selected[start])
new_entity_list[1].append(ent_total_to_selected[end])
new_entity_type.append(edge_type_list[i])
# print(9)
new_entity_list = torch.LongTensor(new_entity_list).to(self.device)
new_entity_type = torch.LongTensor(new_entity_type).to(self.device)
return final_ent_list, ent_neighbors, rel_evt_idxs, evt_neighbors, new_event_index, new_entity_event_index, new_entity_mask, new_entity_list, new_entity_type
def run_epoch(self, epoch, val_mrr = 0):
# print(self.evt2ent_list)
numnum = sum([len(i) for i in self.evt2ent_list])
print("args", numnum)
print("args_type", len(self.role2id))
print("rel_type", len(self.rel2id) // 2)
print("ent_num", self.p.ent_num)
"""
Function to run one epoch of training
Parameters
----------
epoch: current epoch count
Returns