-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathmain_metarh.py
186 lines (165 loc) · 8 KB
/
main_metarh.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
import os
from trainer_metarh import *
from params import *
from data_loader import *
import json
import importlib, sys
importlib.reload(sys)
import itertools, collections
def get_Ground(train_data):
ground = collections.defaultdict(set)
for st in train_data:
ground[tuple(st[:2])].add(st[2])
if len(st) > 3:
qualifier = st[3:]
qualifier_pair = [[qualifier[_*2], qualifier[_*2+1]] for _ in range(len(qualifier)//2)]
for len_ in range(len(qualifier_pair)):
for pairs in itertools.combinations(qualifier_pair, len_+1):
st_ = []
for pair in pairs:
st_ += pair
ground[tuple(st[:2]+st_)].add(st[2])
return ground
def tasks_pad(tasks, max_seq_length):
new_tasks = defaultdict(list)
for rel in tasks:
new_sts = []
for st in tasks[rel]:
new_sts.append(st + ['[PAD]']*(max_seq_length - len(st)))
new_tasks[rel] = new_sts
return new_tasks
def back_pad(back, max_seq_length):
new_back = []
for st in back:
new_back.append(st + ['[PAD]']*(max_seq_length - len(st)))
return new_back
def Ground_pad(Ground, max_seq_length):
new_Ground = defaultdict(list)
for index in Ground:
new_index = list(index) + ['[PAD]']*(max_seq_length-len(index)-1)
new_Ground[tuple(new_index)] = Ground[index]
return new_Ground
if __name__ == '__main__':
params = get_params()
os.environ["CUDA_VISIBLE_DEVICES"] = params['device']
params['device'] = 'cuda:0'
print("---------Parameters---------")
for k, v in params.items():
print(k + ': ' + str(v))
print("----------------------------")
# control random seed
if params['seed'] is not None:
seed = int(params['seed'])
os.environ['PYTHONHASHSEED'] = str(seed)
os.environ["CUBLAS_WORKSPACE_CONFIG"] = ":4096:8"
torch.manual_seed(seed)
torch.cuda.manual_seed(seed)
torch.cuda.manual_seed_all(seed)
torch.backends.cudnn.deterministic = True
torch.backends.cudnn.benchmark = False
torch.backends.cudnn.enabled = False
np.random.seed(seed)
random.seed(seed)
print('random seed {}'.format(seed))
torch.use_deterministic_algorithms(True)
# select the dataset
for k, v in data_dir.items():
data_dir[k] = params['data_path']+params['dataset']+v
dataset = dict()
print("loading train_tasks_in_train ... ...")
print("loading test_tasks ... ...")
print("loading dev_tasks ... ...")
train_tasks = json.load(open(data_dir['train_tasks']))
background = json.load(open(data_dir['background']))
test_tasks = json.load(open(data_dir['test_tasks']))
dev_tasks = json.load(open(data_dir['dev_tasks']))
dataset['rel2candidates'] = json.load(open(data_dir['rel2candidates_in_train']))
train_data, test_data, dev_data = [], [], []
if params['use_in_train']:
train_data = background
for rel in train_tasks:
for st in train_tasks[rel]:
train_data.append(st)
for rel in test_tasks:
for idx, st in enumerate(test_tasks[rel]):
if idx < params['few']:
train_data.append(st)
else:
test_data.append(st)
for rel in dev_tasks:
for idx, st in enumerate(dev_tasks[rel]):
if idx < params['few']:
train_data.append(st)
else:
dev_data.append(st)
Ground = get_Ground(train_data+test_data+dev_data)
Ground_t = get_Ground(train_data)
Ground_t = Ground_pad(Ground_t, params['max_seq_length'])
dataset['Ground_t'] = Ground_t
Ground = Ground_pad(Ground, params['max_seq_length'])
dataset['Ground'] = Ground
new_train_tasks = defaultdict(list)
for rel in train_tasks:
new_train_tasks[rel] = train_tasks[rel]
if params['use_in_train']:
for st in background:
new_train_tasks[st[1]].append(st)
train_tasks = tasks_pad(new_train_tasks, params['max_seq_length'])
test_tasks = tasks_pad(test_tasks, params['max_seq_length'])
dev_tasks = tasks_pad(dev_tasks, params['max_seq_length'])
background = back_pad(background, params['max_seq_length'])
if params['use_in_train']:
all_data = background
else:
all_data = []
for tasks in [train_tasks, test_tasks, dev_tasks]:
for rel in tasks:
all_data += tasks[rel]
for st in all_data:
index = tuple([st[0], st[1]]+st[3:])
assert st[2] in Ground[index]
dataset['train_tasks'] = train_tasks
dataset['dev_tasks'] = dev_tasks
dataset['test_tasks'] = test_tasks
print("----------------------------")
# data_loader
train_data_loader = DataLoader(dataset, params, step='train')
dev_data_loader = DataLoader(dataset, params, step='dev')
test_data_loader = DataLoader(dataset, params, step='test')
data_loaders = [train_data_loader, dev_data_loader, test_data_loader]
# trainer
trainer = Trainer(data_loaders, dataset, params, background)
if params['step'] == 'train':
best_epoch = trainer.train()
print("test")
trainer.reload()
result, bi_result, n_result = trainer.eval(istest=False)
test_result, test_bi_result, test_n_result = trainer.eval_by_relation(istest=True)
print('!'*77)
print("验证集 MRR: {:.3f}\tHits@10: {:.3f}\tHits@5: {:.3f}\tHits@3: {:.3f}\tHits@1: {:.3f}\tBest_epoch: {:.3f}\r".format(
result['MRR'], result['Hits@10'], result['Hits@5'], result['Hits@3'], result['Hits@1'], best_epoch))
print("验证集 bi MRR: {:.3f}\tHits@10: {:.3f}\tHits@5: {:.3f}\tHits@3: {:.3f}\tHits@1: {:.3f}\tBest_epoch: {:.3f}\r".format(
bi_result['MRR'], bi_result['Hits@10'], bi_result['Hits@5'], bi_result['Hits@3'], bi_result['Hits@1'], best_epoch))
print("验证集 n MRR: {:.3f}\tHits@10: {:.3f}\tHits@5: {:.3f}\tHits@3: {:.3f}\tHits@1: {:.3f}\tBest_epoch: {:.3f}\r".format(
n_result['MRR'], n_result['Hits@10'], n_result['Hits@5'], n_result['Hits@3'], n_result['Hits@1'], best_epoch))
print("测试集 MRR: {:.3f}\tHits@10: {:.3f}\tHits@5: {:.3f}\tHits@3: {:.3f}\tHits@1: {:.3f}\tBest_epoch: {:.3f}\r".format(
test_result['MRR'], test_result['Hits@10'], test_result['Hits@5'], test_result['Hits@3'], test_result['Hits@1'], best_epoch))
print("测试集 bi MRR: {:.3f}\tHits@10: {:.3f}\tHits@5: {:.3f}\tHits@3: {:.3f}\tHits@1: {:.3f}\tBest_epoch: {:.3f}\r".format(
test_bi_result['MRR'], test_bi_result['Hits@10'], test_bi_result['Hits@5'], test_bi_result['Hits@3'], test_bi_result['Hits@1'], best_epoch))
print("测试集 n MRR: {:.3f}\tHits@10: {:.3f}\tHits@5: {:.3f}\tHits@3: {:.3f}\tHits@1: {:.3f}\tBest_epoch: {:.3f}\r".format(
test_n_result['MRR'], test_n_result['Hits@10'], test_n_result['Hits@5'], test_n_result['Hits@3'], test_n_result['Hits@1'], best_epoch))
elif params['step'] == 'test':
trainer.reload()
test_result, test_bi_result, test_n_result = trainer.eval_by_relation(istest=True)
print("测试集 MRR: {:.3f}\tHits@10: {:.3f}\tHits@5: {:.3f}\tHits@3: {:.3f}\tHits@1: {:.3f}\r".format(
test_result['MRR'], test_result['Hits@10'], test_result['Hits@5'], test_result['Hits@3'], test_result['Hits@1']))
elif params['step'] == 'dev':
trainer.reload()
result, test_bi_result, test_n_result = trainer.eval_by_relation(istest=False)
print("验证集 MRR: {:.3f}\tHits@10: {:.3f}\tHits@5: {:.3f}\tHits@3: {:.3f}\tHits@1: {:.3f}\r".format(
result['MRR'], result['Hits@10'], result['Hits@5'], result['Hits@3'], result['Hits@1']))
elif params['step'] == 'case':
trainer.reload()
result = trainer.case_eval(istest=True)
print("验证集 MRR: {:.3f}\tHits@10: {:.3f}\tHits@5: {:.3f}\tHits@3: {:.3f}\tHits@1: {:.3f}\r".format(
result['MRR'], result['Hits@10'], result['Hits@5'], result['Hits@3'], result['Hits@1']))