-
Notifications
You must be signed in to change notification settings - Fork 2
/
neste_d.py
190 lines (163 loc) · 7.41 KB
/
neste_d.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
import openke
from openke.config import Trainer, Tester
from openke.module.model import NestE_D
from openke.module.loss import SoftplusLoss
from openke.module.strategy import NegativeSampling
from openke.data import TrainDataLoader, TestDataLoader
import argparse
import os
parser = argparse.ArgumentParser(description="Hyperparameters")
parser.add_argument('data')
parser.add_argument('alpha', type=float, help="learning rate")
parser.add_argument('regul', type=float, help="regul_rate")
parser.add_argument('epoch', type=int, help="epoch")
parser.add_argument('--dim', type=int, default=200)
parser.add_argument('--meta', type=float, default=0.0, help="weight_meta")
parser.add_argument('--aug', type=float, default=0.0, help="weight_aug")
parser.add_argument('--lp', default=False, action="store_true")
parser.add_argument('--tp', default=False, action="store_true")
parser.add_argument('--clp', default=False, action="store_true")
parser.add_argument('--serial', type=int, default=0)
args = parser.parse_args()
### SET RANDOM SEED ###
import torch
import numpy as np
import random
random.seed(args.serial)
np.random.seed(args.serial)
torch.manual_seed(args.serial)
OMP_NUM_THREADS=8
torch.set_num_threads(8)
### PRINT INFO ###
print("data: {} alpha: {} regul: {} meta: {} aug: {} epoch: {}".format(args.data, args.alpha, args.regul, args.meta, args.aug, args.epoch))
# dataloader for training
train_dataloader = TrainDataLoader(
in_path="./benchmarks/{}/base/".format(args.data),
nbatches=100,
threads=8,
sampling_mode="normal",
bern_flag=1,
filter_flag=1,
neg_ent=25,
neg_rel=0
)
# Prepare file for meta
meta_rel_tot = 0
list_entity_meta = []
list_meta = []
batch_meta_size = 0
with open("./benchmarks/{}/meta/relation2id.txt".format(args.data), 'r') as f:
meta_rel_tot = int(f.readline().strip())
if args.meta != 0.0:
with open("./benchmarks/{}/meta/entity2id.txt".format(args.data), 'r') as f:
f.readline()
for line in f.readlines():
triplet, _ = line.strip().split()
h, t, r = triplet.strip().split('_')
list_entity_meta.append([h, t, r])
list_entity_meta = np.array(list_entity_meta).astype(int)
with open("./benchmarks/{}/meta/train2id.txt".format(args.data), 'r') as f:
f.readline()
for line in f.readlines():
h, t, r = line.strip().split()
list_meta.append([h, t, r])
list_meta = np.array(list_meta).astype(int)
batch_meta_size = int(len(list_meta) / 100)
# Prepare file for aug
list_aug = []
batch_aug_size = 0
if args.aug != 0.0:
with open("./benchmarks/{}/base/augment2id.txt".format(args.data), 'r') as f:
f.readline()
for line in f.readlines():
h, t, r = line.strip().split()
list_aug.append([h, t, r])
list_aug = np.array(list_aug).astype(int)
batch_aug_size = int(len(list_aug) / 100)
# define the model
NestE_D = NestE_D(
ent_tot=train_dataloader.get_ent_tot(),
rel_tot=train_dataloader.get_rel_tot(),
meta_rel_tot=meta_rel_tot,
dim=args.dim,
seed=args.serial
)
try:
print("Loading checkpoint alpha_{}_regul_{}_meta_{}_aug_{}_epoch_{}-{}.ckpt".format(args.alpha, args.regul, args.meta, args.aug, args.epoch, args.serial))
NestE_D.load_checkpoint('./checkpoint/{}/NestE_D/alpha_{}_regul_{}_meta_{}_aug_{}_epoch_{}-{}.ckpt'.format(args.data, args.alpha, args.regul, args.meta, args.aug, args.epoch, args.serial))
except:
print("No checkpoint available!")
# define the loss function
model = NegativeSampling(
model=NestE_D,
loss=SoftplusLoss(),
batch_size=train_dataloader.get_batch_size(),
regul_rate=args.regul,
meta=(args.meta != 0.0),
batch_meta_size=batch_meta_size,
aug=(args.aug != 0.0),
batch_aug_size=batch_aug_size
)
# train the model
trainer = Trainer(
model=model,
data_loader=train_dataloader,
train_times=args.epoch,
alpha=args.alpha,
use_gpu=True,
opt_method="adagrad",
list_entity_meta=list_entity_meta,
list_meta=list_meta,
batch_meta_size=batch_meta_size,
weight_meta=args.meta,
list_aug=list_aug,
batch_aug_size=batch_aug_size,
weight_aug=args.aug
)
trainer.run()
os.makedirs('./checkpoint/{}/NestE_D'.format(args.data), exist_ok=True)
NestE_D.save_checkpoint('./checkpoint/{}/NestE_D/alpha_{}_regul_{}_meta_{}_aug_{}_epoch_{}-{}.ckpt'.format(args.data, args.alpha, args.regul, args.meta, args.aug, args.epoch, args.serial))
if args.lp:
test_dataloader = TestDataLoader("./benchmarks/{}/base/".format(args.data), "link")
tester = Tester(model=NestE_D, data_loader=test_dataloader, use_gpu=True)
mr, mrr, hit10, hit3, hit1 = tester.run_link_prediction(type_constrain=False)
elif args.tp:
test_dataloader = TestDataLoader("./benchmarks/{}/meta/".format(args.data), "link")
tester = Tester(model=NestE_D, data_loader=test_dataloader, use_gpu=True)
mr, mrr, hit10, hit3, hit1 = tester.run_triplet_prediction(type_constrain=False, list_entity_meta=list_entity_meta)
elif args.clp:
list_info = []
with open("./benchmarks/{}/conditional/relation2id.txt".format(args.data), 'r') as f:
f.readline()
for line in f.readlines():
relationship, _ = line.strip().split()
h1, t1, r1, R, r2 = relationship.split('_')
list_info.append((h1, t1, r1, R, r2))
list_info = np.array(list_info).astype(int)
test_dataloader = TestDataLoader("./benchmarks/{}/conditional/".format(args.data), "link")
tester = Tester(model=NestE_D, data_loader=test_dataloader, use_gpu=True)
mr_tail, mrr_tail, hit10_tail, hit3_tail, hit1_tail = tester.run_conditional_link_prediction(type_constrain=False, list_info=list_info, weight_meta=args.meta)
list_info = []
with open("./benchmarks/{}/conditional_head/relation2id.txt".format(args.data), 'r') as f:
f.readline()
for line in f.readlines():
relationship, _ = line.strip().split()
r1, R, h2, t2, r2 = relationship.split('_')
list_info.append((r1, R, h2, t2, r2))
list_info = np.array(list_info).astype(int)
test_dataloader = TestDataLoader("./benchmarks/{}/conditional_head/".format(args.data), "link")
tester = Tester(model=NestE_D, data_loader=test_dataloader, use_gpu=True)
mr_head, mrr_head, hit10_head, hit3_head, hit1_head = tester.run_conditional_link_prediction_head(type_constrain=False, list_info=list_info, weight_meta=args.meta)
mr, mrr, hit10, hit3, hit1 = (mr_tail + mr_head) / 2, (mrr_tail + mrr_head) / 2, (hit10_tail + hit10_head) / 2, (hit3_tail + hit3_head) / 2, (hit1_tail + hit1_head) / 2
if args.lp:
os.makedirs("./result_test/{}/NestE_D/LP".format(args.data), exist_ok=True)
with open("./result_test/{}/NestE_D/LP/alpha_{}_regul_{}_meta_{}_aug_{}_epoch_{}-{}.txt".format(args.data, args.alpha, args.regul, args.meta, args.aug, args.epoch, args.serial), 'w') as f:
f.write("LP: {} {} {} {} {}\n".format(mr, mrr, hit10, hit3, hit1))
elif args.tp:
os.makedirs("./result_test/{}/NestE_D/TP".format(args.data), exist_ok=True)
with open("./result_test/{}/NestE_D/TP/alpha_{}_regul_{}_meta_{}_aug_{}_epoch_{}-{}.txt".format(args.data, args.alpha, args.regul, args.meta, args.aug, args.epoch, args.serial), 'w') as f:
f.write("TP: {} {} {} {} {}\n".format(mr, mrr, hit10, hit3, hit1))
elif args.clp:
os.makedirs("./result_test/{}/NestE_D/CLP".format(args.data), exist_ok=True)
with open("./result_test/{}/NestE_D/CLP/alpha_{}_regul_{}_meta_{}_aug_{}_epoch_{}-{}.txt".format(args.data, args.alpha, args.regul, args.meta, args.aug, args.epoch, args.serial), 'w') as f:
f.write("CLP: {} {} {} {} {}\n".format(mr, mrr, hit10, hit3, hit1))