-
Notifications
You must be signed in to change notification settings - Fork 5
/
n_clo.py
187 lines (153 loc) · 7.03 KB
/
n_clo.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
import pandas as pd
from n_model import CloModel
import os
import pandas as pd
import torch
from tqdm import tqdm
from n_model import HCL
import dgl
import torch
import time
from sklearn.metrics import precision_recall_fscore_support
import argparse
from n_parse_code import clo_datasetSplit, DataSet
if __name__ == '__main__':
parser = argparse.ArgumentParser(description="Code Clone Detection")
parser.add_argument('--pair_path', default='data/clone_pair/oj_clone_ids.pkl',
help='The path of the clone pairs')
parser.add_argument('--data_path', default='data/oj/programs.pkl',
help='data path')
parser.add_argument('--pre_model', default='model.pkl',
help='pre_model')
parser.add_argument('--device', default='cpu',
help='device')
args = parser.parse_args()
'''
变量进行保存
'''
pair_path = args.pair_path
data_path = args.data_path
device = args.device
'''
根据数据集需要更改的信息
'''
dataset_name = 'oj'
train_ratio, val_ratio, test_ratio = 8, 1, 1
train_data1, train_data2, train_data_label, valid_data1, valid_data2, valid_data_label, test_data1, test_data2, test_data_label \
= clo_datasetSplit(pair_path,data_path,train_ratio,val_ratio,test_ratio)
begin = 0
lr, BATCH_SIZE, EPOCH = 0.00001, 2, 5
USE_GPU = False
in_feats, n_layer, n_head, drop_out, n_class = 768, 4, 4, 0.5, 1
# pre_model = torch.load(args.pre_model)
# pre_model_dict = pre_model.state_dict()
model = CloModel(in_feats, n_layer, drop_out, n_class, device='cpu').to(device)
# model_dict = model.state_dict()
# pretrained_dict = {k: v for k, v in pre_model_dict.items() if k in model_dict}
# model_dict.update(pretrained_dict)
# model.load_state_dict(model_dict)
optimizer = torch.optim.Adam(model.parameters(), lr=lr)
loss_function = torch.nn.BCELoss()
precision, recall, f1 = 0, 0, 0
pbar = tqdm(range(EPOCH))
best_model = model
best_acc = 0.0
train_loss_ = []
val_loss_ = []
train_acc_ = []
val_acc_ = []
dataSet = DataSet()
for t in range(1, n_class + 1):
train_data1_t, train_data2_t, test_data1_t, test_data2_t, train_label_t, test_label_t = [], [], [], [], [], []
if dataset_name == 'bcb':
for i in range(len(train_data_label)):
if train_data_label == t:
train_data1_t.append(train_data1[i])
train_data2_t.append(train_data2[i])
train_label_t.append[1]
for i in range(len(test_data_label)):
if test_data_label[i] == t:
test_data1_t.append(test_data1[i])
test_data2_t.append(test_data2[i])
test_label_t.append[1]
else:
train_data1_t, train_data2_t, test_data1_t, test_data2_t = train_data1, train_data2, test_data1, test_data2
train_label_t,test_label_t = train_data_label, test_data_label
for epoch in pbar:
pbar.set_description('epoch:%d processing' % (epoch))
i = 0
start_time = time.time()
total_acc = 0.0
total_loss = 0.0
total = 0.0
model.train()
while (i + BATCH_SIZE) <= len(train_data1_t):
model.train()
input_node1, path1, label1, p_n_dict1, n_ast_node1, src1, dst1, downtown_label1 = dataSet.parse_c(
train_data1.iloc[i:i + BATCH_SIZE, :], dataset_name)
input_node2, path2, label2, p_n_dict2, n_ast_node2, src2, dst2, downtown_label2 = dataSet.parse_c(
train_data2.iloc[i:i + BATCH_SIZE, :], dataset_name)
label_batch = train_label_t[i:i+BATCH_SIZE]
i = i + BATCH_SIZE
'''
输入数据类型转换，并选择需要的运行环境
'''
g1 = dgl.graph((src1, dst1))
g1 = dgl.add_self_loop(g1).to(device)
g2 = dgl.graph((src2, dst2))
g2 = dgl.add_self_loop(g2).to(device)
node_emb1 = torch.FloatTensor(input_node1).to(device)
path_emb1 = torch.FloatTensor(path1).to(device)
node_emb2 = torch.FloatTensor(input_node2).to(device)
path_emb2 = torch.FloatTensor(path2).to(device)
labels = torch.FloatTensor(label_batch).to(device)
model.zero_grad()
logits = model(g1, node_emb1, n_ast_node1, path_emb1, p_n_dict1, g2, node_emb2, n_ast_node2, path_emb2, p_n_dict2)
labels = labels.view(-1, 1)
print('n_clo',labels.shape)
loss = loss_function(logits, labels)
loss.backward()
optimizer.step()
print("Testing-%d...")
# testing procedure
predicts = []
trues = []
total_loss = 0.0
total = 0.0
i = 0
while (i + BATCH_SIZE) < len(test_data1_t):
input_node1, path1, label1, p_n_dict1, n_ast_node1, src1, dst1, downtown_label1 = dataSet.parse_c(
test_data1.iloc[i:i + BATCH_SIZE, :], dataset_name)
input_node2, path2, label2, p_n_dict2, n_ast_node2, src2, dst2, downtown_label2 = dataSet.parse_c(
test_data2.iloc[i:i + BATCH_SIZE, :], dataset_name)
label_batch = test_label_t[i:i + BATCH_SIZE]
i = i + BATCH_SIZE
g1 = dgl.graph((src1, dst1))
g1 = dgl.add_self_loop(g1).to(device)
g2 = dgl.graph((src2, dst2))
g2 = dgl.add_self_loop(g2).to(device)
node_emb1 = torch.FloatTensor(input_node1).to(device)
path_emb1 = torch.FloatTensor(path1).to(device)
node_emb2 = torch.FloatTensor(input_node2).to(device)
path_emb2 = torch.FloatTensor(path2).to(device)
labels = torch.FloatTensor(label_batch).to(device)
model.zero_grad()
logits = model(g1, node_emb1, n_ast_node1, path_emb1, p_n_dict1, g2, node_emb2, n_ast_node2, path_emb2, p_n_dict2)
labels = labels.view(-1, 1)
loss = loss_function(logits, labels)
# calc testing acc
predicted = (logits.data > 0.5).cpu().numpy()
predicts.extend(predicted)
trues.extend(labels.cpu().numpy())
total += len(labels)
total_loss += loss.item() * len(labels)
if dataset_name == 'bcb':
weights = [0, 0.005, 0.001, 0.002, 0.010, 0.982]
p, r, f, _ = precision_recall_fscore_support(trues, predicts, average='binary')
precision += weights[t] * p
recall += weights[t] * r
f1 += weights[t] * f
print("Type-" + str(t) + ": " + str(p) + " " + str(r) + " " + str(f))
else:
precision, recall, f1, _ = precision_recall_fscore_support(trues, predicts, average='binary')
print("Total testing results(P,R,F1):%.3f, %.3f, %.3f" % (precision, recall, f1))