/
case_study.py
175 lines (149 loc) · 6.8 KB
/
case_study.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
# coding=UTF-8
import os
import numpy as np
import random
import argparse
import torch
from core.sim_metric import EDSim, SDPSim, CosSim
from core.learn_entry import fine_tune, pretrain_encoder
from core.dataloader import load_graph, load_seed
from core.generator import Generator
from core.discriminator import Discriminator
from core.evaluate import evaluate_generator
from core.learn_multi_view import update_s
SIM_TABLE = {
'ed': EDSim(),
'sdp': SDPSim(),
'cos': CosSim()
}
def regist_parser():
parser = argparse.ArgumentParser()
parser.add_argument('--dataset', type=str, default='data/CoNLL')
parser.add_argument('--input_model_file', type=str, default='')
parser.add_argument('--output_model_file', type=str, default='')
parser.add_argument('--method', type=str, default='multi_view')
parser.add_argument('--sim_metric', type=str, default='cos',
help='sim metric function,choose from list ['
'"ed"(euclidean distance),"sdp"(scaled dot product)'
',"cos"(cosine)]')
parser.add_argument('--n_iter', type=int, default=20)
parser.add_argument('--min_match', type=int, default=1)
parser.add_argument('--n_expansion', type=int, default=10)
parser.add_argument('--k_hop', type=int, default=2)
parser.add_argument('--local', action='store_true')
parser.add_argument('--mean_updated', action='store_true')
parser.add_argument('--n_layer', type=int, default=3)
parser.add_argument('--dropout', type=float, default=0.1)
parser.add_argument('--bias', action='store_true')
parser.add_argument('--device', type=int, default=4)
parser.add_argument('--cpu', action='store_true', help='Ignore CUDA.')
parser.add_argument('--seed', type=int, default=1)
parser.add_argument('--gamma', type=float, default=1.0)
parser.add_argument('--init_encoder_epoch', type=int, default=100)
parser.add_argument('--init_decoder_epoch', type=int, default=100)
parser.add_argument('--encoder_epoch', type=int, default=50)
parser.add_argument('--decoder_epoch', type=int, default=50)
parser.add_argument('--optimizer', default='adam', help='Optimizer')
parser.add_argument('--lr', type=float, default=1e-4, help='Learning rate')
parser.add_argument('--decay', type=float, default=1e-3,
help='Weight decay for optimization')
parser.add_argument('--feature_type', default='glove')
parser.add_argument('--feature_dim', type=int, default=50)
parser.add_argument('--edge_feature_dim', type=int, default=5)
args = parser.parse_args()
return args
def load_tables(file_path):
table = []
with open(file_path, 'r') as f:
for i, line in enumerate(f):
line = line.strip().split('\t')
if len(line) > 1:
table.append(line[0] + '(' + line[1] + ')')
else:
table.append(line[0])
return table
def print_case(name, model, graph, graph_data, seeds,
entity_table, label_table):
_, expansions, _ = model(graph_data, seeds, 20)
n = len(label_table)
errors = [[0 for _ in range(n)] for _ in range(n)]
with open(name+'.tsv', 'w') as f:
for it, _expansions in enumerate(expansions):
f.write('=======Iteration %d=======\t \t \n' % it)
for i, expansion in enumerate(_expansions):
original_i = int(graph.node_s.itol[i])
original_label = label_table[original_i]
r_set = set()
f_set = set()
for en in expansion.cpu().tolist():
original_en = entity_table[int(graph.node_s.itos[en])]
if graph_data.m_y[en][i] == 1:
r_set.add(original_en)
else:
f_set.add(original_en)
true_i = int(graph.node_s.itol[graph_data.y[en].cpu().item()])
errors[true_i][original_i] += 1
f.write(str(original_label) + '\t' + ','.join(r_set) + '\t' + ','.join(f_set) + '\n')
f.write('\n')
f.write('\n==========errors==========\n')
f.write(' OOOOO ')
for i in range(n):
f.write('\t'+label_table[i])
f.write('\n')
total = 0
for i in range(n):
f.write(label_table[i] + '\t' + '\t'.join(list(map(str, errors[i]))))
total += sum(errors[i])
f.write('\n')
f.write('TOTAL ERRORS: %d'%total)
if __name__ == '__main__':
args = regist_parser()
if args.seed:
os.environ['PYTHONHASHSEED'] = str(args.seed)
torch.manual_seed(args.seed)
np.random.seed(args.seed)
torch.cuda.manual_seed(args.seed)
torch.cuda.manual_seed_all(args.seed)
torch.backends.cudnn.deterministic = True
torch.backends.cudnn.benchmark = False
# cuda device setting
if args.cpu and args.device is not None or not torch.cuda.is_available():
args.device = torch.device('cpu')
else:
args.device = torch.device('cuda:%d' % args.device)
opt = vars(args)
if opt['local']:
opt['k_hop'] = 1
if opt['feature_type'] == 'bert':
opt['feature_dim'] = 768
device = opt['device']
opt['sim_metric'] = SIM_TABLE[opt['sim_metric'].lower()].to(device)
pkl_path = 'graph_' + opt['feature_type'] + '.pkl'
graph_data, graph = load_graph(opt, pkl_path, pyg=True)
seed_file = opt['dataset'] + '/seeds.txt'
seeds = load_seed(graph.node_s, seed_file)
seeds = [torch.LongTensor(seed) for seed in seeds]
opt['n_class'] = len(graph.node_s.itol)
generator = Generator(opt)
graph_data = graph_data.to(opt['device'])
graph_data.x = (graph_data.x[0].to(opt['device']),
graph_data.x[1].to(opt['device']))
seeds = [seed.to(opt['device']) for seed in seeds]
generator = generator.to(opt['device'])
entity_table = load_tables(opt['dataset'] + '/entities.txt')
label_table = load_tables(opt['dataset'] + '/labels.txt')
classifier = pretrain_encoder(opt, generator.encoder, graph_data, seeds)
update_s(opt, generator, classifier, graph_data, seeds)
evaluate_generator(generator, graph_data, seeds, n_iter=20)
dataset_name = os.path.split(opt['dataset'])[1].lower()
print_case('results/original_' + dataset_name, generator, graph,
graph_data, seeds, entity_table, label_table)
print('++++++adversarial learning++++++++')
if opt['input_model_file']:
pkl_path = opt['input_model_file']
generator = generator.cpu()
generator.load_state_dict(torch.load(pkl_path))
generator = generator.to(opt['device'])
evaluate_generator(generator, graph_data, seeds, n_iter=20)
print_case('results/' + dataset_name, generator, graph, graph_data, seeds,
entity_table, label_table)