-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathrun.py
181 lines (159 loc) · 7.47 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
import argparse
from dgl.dataloading import NodeDataLoader, MultiLayerFullNeighborSampler, MultiLayerNeighborSampler
from utils import *
from MCLET import MCLET
def main(args):
use_cuda = args['cuda'] and torch.cuda.is_available()
data_path = os.path.join(args['data_dir'], args['dataset'])
save_path = os.path.join(args['save_dir'], args['dataset'], args['save_name'])
# graph
e2id = read_id(os.path.join(data_path, 'entities.tsv'))
r2id = read_id(os.path.join(data_path, 'relations.tsv'))
r2id['type'] = len(r2id)
r2id['type_cluster'] = len(r2id)
r2id['entity_cluster'] = len(r2id)
t2id = read_id(os.path.join(data_path, 'types.tsv'))
c2id = read_id(os.path.join(data_path, 'clusters.tsv'))
num_entities = len(e2id)
num_rels = len(r2id)
num_types = len(t2id)
num_clusters = len(c2id)
g, train_label, all_true, train_id, valid_id, test_id = load_graph(data_path, e2id, r2id, t2id, c2id, args['load_ET'], args['load_KG'], args['load_TC'], args['load_EC'])
e2t_graph, t2c_graph, e2c_graph = build_e2t_graph(args, e2id, t2id), build_t2c_graph(args, t2id, c2id), build_e2c_graph(args, e2id, c2id)
if args['neighbor_sampling']:
train_sampler = MultiLayerNeighborSampler([args['neighbor_num']] * args['num_layers'], replace=True)
else:
train_sampler = MultiLayerFullNeighborSampler(args['num_layers'])
test_sampler = MultiLayerFullNeighborSampler(args['num_layers'])
train_dataloader = NodeDataLoader(
g, train_id, train_sampler,
batch_size=args['train_batch_size'],
shuffle=True,
drop_last=False,
num_workers=6
)
valid_dataloader = NodeDataLoader(
g, valid_id, test_sampler,
batch_size=args['test_batch_size'],
shuffle=False,
drop_last=False,
num_workers=6
)
# model
model = MCLET(args, num_entities, num_rels, num_types, num_clusters, e2t_graph=e2t_graph, t2c_graph=t2c_graph, e2c_graph=e2c_graph)
if use_cuda:
model = model.to('cuda')
for name, param in model.named_parameters():
logging.debug('Parameter %s: %s, require_grad=%s' % (name, str(param.size()), str(param.requires_grad)))
# optimizer
optimizer = torch.optim.Adam(
filter(lambda p: p.requires_grad, model.parameters()),
lr=args['lr'],
)
# loss
criterion = torch.nn.BCELoss()
# training
max_valid_mrr = 0
model.train()
for epoch in range(args['max_epoch']):
log = []
for input_nodes, output_nodes, blocks in train_dataloader:
label = train_label[output_nodes, :]
if use_cuda:
blocks = [b.to(torch.device('cuda')) for b in blocks]
label = label.cuda()
predict, aux_loss = model(blocks)
if args['loss'] == 'BCE':
loss = criterion(predict, label) + aux_loss
pos_loss, neg_loss = loss, loss
elif args['loss'] == 'FNA':
pos_loss, neg_loss = cal_loss(predict, label, args['beta'])
loss = pos_loss + neg_loss + aux_loss
else:
raise ValueError('loss %s is not defined' % args['loss'])
log.append({
"loss": loss.item(),
"pos_loss": pos_loss.item(),
"neg_loss": neg_loss.item(),
})
optimizer.zero_grad()
loss.backward()
optimizer.step()
avg_loss = sum([_['loss'] for _ in log]) / len(log)
avg_pos_loss = sum([_['pos_loss'] for _ in log]) / len(log)
avg_neg_loss = sum([_['neg_loss'] for _ in log]) / len(log)
logging.debug('epoch %d: loss: %f\tpos_loss: %f\tneg_loss: %f' % (epoch, avg_loss, avg_pos_loss, avg_neg_loss))
if epoch != 0 and epoch % args['valid_epoch'] == 0:
torch.save(model.state_dict(), os.path.join(save_path, 'model.pkl'))
model.eval()
with torch.no_grad():
predict = torch.zeros(num_entities, num_types, dtype=torch.half)
for input_nodes, output_nodes, blocks in valid_dataloader:
if use_cuda:
blocks = [b.to(torch.device('cuda')) for b in blocks]
predict_result, aux_loss = model(blocks)
predict[output_nodes] = predict_result.cpu().half()
valid_mrr = evaluate(os.path.join(data_path, 'ET_valid.txt'), predict, all_true, e2id, t2id)
model.train()
if valid_mrr < max_valid_mrr:
logging.debug('early stop')
break
else:
torch.save(model.state_dict(), os.path.join(save_path, 'best_model.pkl'))
max_valid_mrr = valid_mrr
with torch.no_grad():
model.load_state_dict(torch.load(os.path.join(save_path, 'best_model.pkl')))
model.eval()
predict = torch.zeros(num_entities, num_types, dtype=torch.half)
test_dataloader = NodeDataLoader(
g, test_id, test_sampler,
batch_size=args['test_batch_size'],
shuffle=False,
drop_last=False,
num_workers=6
)
for input_nodes, output_nodes, blocks in test_dataloader:
if use_cuda:
blocks = [b.to(torch.device('cuda')) for b in blocks]
predict_result, aux_loss = model(blocks)
predict[output_nodes] = predict_result.cpu().half()
evaluate(os.path.join(data_path, 'ET_test.txt'), predict, all_true, e2id, t2id)
def get_params():
parser = argparse.ArgumentParser()
parser.add_argument('--data_dir', type=str, default='./data')
parser.add_argument('--dataset', type=str, default='FB15kET')
parser.add_argument('--load_ET', action='store_true', default=True)
parser.add_argument('--load_KG', action='store_true', default=True)
parser.add_argument('--load_TC', action='store_true', default=True)
parser.add_argument('--load_EC', action='store_true', default=True)
parser.add_argument('--neighbor_sampling', action='store_true', default=True)
parser.add_argument('--save_dir', type=str, default='save')
parser.add_argument('--save_name', type=str, default='save_name')
parser.add_argument('--hidden_dim', type=int, default=100)
parser.add_argument('--lr', type=float, default=0.001)
parser.add_argument('--neighbor_num', type=int, default=10)
parser.add_argument('--train_batch_size', type=int, default=128)
parser.add_argument('--test_batch_size', type=int, default=64)
parser.add_argument('--cuda', action='store_true', default=True)
parser.add_argument('--max_epoch', type=int, default=1000)
parser.add_argument('--valid_epoch', type=int, default=25)
parser.add_argument('--beta', type=float, default=4.0)
parser.add_argument('--loss', type=str, default='FNA')
parser.add_argument('--num_layers', type=int, default=1)
parser.add_argument('--decay', type=float, default=1e-5)
parser.add_argument('--embedding_dropout', type=float, default=0.3)
# ablation params
parser.add_argument('--lightgcn_layer', type=int, default=2)
parser.add_argument('--cl_temperature', type=float, default=0.6)
parser.add_argument('--cl_loss_weight', type=float, default=0.001)
parser.add_argument('--num_heads', type=int, default=5)
args, _ = parser.parse_known_args()
return args
if __name__ == '__main__':
try:
params = vars(get_params())
set_logger(params)
main(params)
except Exception as e:
logging.exception(e)
raise