-
Notifications
You must be signed in to change notification settings - Fork 20
/
Copy pathvisualize_bcb.py
188 lines (182 loc) · 6.97 KB
/
visualize_bcb.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
import math
import random
import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.autograd import Variable
import torch.optim as optim
import numpy as np
import time
import sys
import argparse
from tqdm import tqdm, trange
import pycparser
from createclone_bcb import createast,creategmndata,createseparategraph
import models
from torch_geometric.data import Data, DataLoader
parser = argparse.ArgumentParser()
parser.add_argument("--cuda", default=True)
parser.add_argument("--dataset", default='gcj')
parser.add_argument("--graphmode", default='astandnext')
parser.add_argument("--nextsib", default=False)
parser.add_argument("--ifedge", default=False)
parser.add_argument("--whileedge", default=False)
parser.add_argument("--foredge", default=False)
parser.add_argument("--blockedge", default=False)
parser.add_argument("--nexttoken", default=False)
parser.add_argument("--nextuse", default=False)
parser.add_argument("--data_setting", default='11')
parser.add_argument("--batch_size", default=32)
parser.add_argument("--num_layers", default=4)
parser.add_argument("--num_epochs", default=4)
parser.add_argument("--lr", default=0.001)
parser.add_argument("--threshold", default=0)
args = parser.parse_args()
device=torch.device('cuda:0')
#device=torch.device('cpu')
astdict,vocablen,vocabdict=createast()
treedict=createseparategraph(astdict, vocablen, vocabdict,device,mode=args.graphmode,nextsib=args.nextsib,ifedge=args.ifedge,whileedge=args.whileedge,foredge=args.foredge,blockedge=args.blockedge,nexttoken=args.nexttoken,nextuse=args.nextuse)
traindata,validdata,testdata=creategmndata(args.data_setting,treedict,vocablen,vocabdict,device)
print(len(traindata))
#trainloder=DataLoader(traindata,batch_size=1)
num_layers=int(args.num_layers)
model=models.GMNnet(vocablen,embedding_dim=100,num_layers=num_layers,device=device).to(device)
optimizer = optim.Adam(model.parameters(), lr=args.lr)
criterion=nn.CosineEmbeddingLoss()
criterion2=nn.MSELoss()
#model=torch.load('bcbresult/astandnext_epoch_2').to(device)
#testsample=testdata[113]
#print(testsample[1])
#data=testsample[0]
#label=testsample[1]
#quit()
def visualize(data,label,mode=''):
label=torch.tensor(label, dtype=torch.float, device=device)
x1, x2, edge_index1, edge_index2, edge_attr1, edge_attr2=data
x1=torch.tensor(x1, dtype=torch.long, device=device)
x2=torch.tensor(x2, dtype=torch.long, device=device)
edge_index1=torch.tensor(edge_index1, dtype=torch.long, device=device)
edge_index2=torch.tensor(edge_index2, dtype=torch.long, device=device)
if edge_attr1!=None:
edge_attr1=torch.tensor(edge_attr1, dtype=torch.long, device=device)
edge_attr2=torch.tensor(edge_attr2, dtype=torch.long, device=device)
data=[x1, x2, edge_index1, edge_index2, edge_attr1, edge_attr2]
prediction=model.forward(data,mode=mode)
output=F.cosine_similarity(prediction[0],prediction[1])
prediction = torch.sign(output).item()
print(output)
'''visualize(data,label)
a1=torch.load('attns/a1')
print(a1)
quit()'''
vset=[testdata[41],testdata[81],testdata[113],testdata[115],testdata[131]]
'''for data,label in vset:
print(label)
visualize(data,label)
quit()'''
def create_batches(data):
#random.shuffle(data)
batches = [data[graph:graph+args.batch_size] for graph in range(0, len(data), args.batch_size)]
return batches
def test(dataset):
#model.eval()
count=0
correct=0
tp = 0
tn = 0
fp = 0
fn = 0
results=[]
for data,label in dataset:
label=torch.tensor(label, dtype=torch.float, device=device)
x1, x2, edge_index1, edge_index2, edge_attr1, edge_attr2=data
x1=torch.tensor(x1, dtype=torch.long, device=device)
x2=torch.tensor(x2, dtype=torch.long, device=device)
edge_index1=torch.tensor(edge_index1, dtype=torch.long, device=device)
edge_index2=torch.tensor(edge_index2, dtype=torch.long, device=device)
if edge_attr1!=None:
edge_attr1=torch.tensor(edge_attr1, dtype=torch.long, device=device)
edge_attr2=torch.tensor(edge_attr2, dtype=torch.long, device=device)
data=[x1, x2, edge_index1, edge_index2, edge_attr1, edge_attr2]
prediction=model(data)
output=F.cosine_similarity(prediction[0],prediction[1])
results.append(output.item())
prediction = torch.sign(output).item()
if prediction>args.threshold and label.item()==1:
tp+=1
#print('tp')
if prediction<=args.threshold and label.item()==-1:
tn+=1
#print('tn')
if prediction>args.threshold and label.item()==-1:
fp+=1
#print('fp')
if prediction<=args.threshold and label.item()==1:
fn+=1
#print('fn')
print(tp,tn,fp,fn)
p=0.0
r=0.0
f1=0.0
if tp+fp==0:
print('precision is none')
return
p=tp/(tp+fp)
if tp+fn==0:
print('recall is none')
return
r=tp/(tp+fn)
f1=2*p*r/(p+r)
print('precision')
print(p)
print('recall')
print(r)
print('F1')
print(f1)
return results
epochs = trange(args.num_epochs, leave=True, desc = "Epoch")
for epoch in epochs:# without batching
print('epoch: ',epoch)
batches=create_batches(traindata)
totalloss=0.0
main_index=0.0
for index, batch in tqdm(enumerate(batches), total=len(batches), desc = "Batches"):
optimizer.zero_grad()
batchloss= 0
for data,label in batch:
label=torch.tensor(label, dtype=torch.float, device=device)
x1, x2, edge_index1, edge_index2, edge_attr1, edge_attr2=data
x1=torch.tensor(x1, dtype=torch.long, device=device)
x2=torch.tensor(x2, dtype=torch.long, device=device)
edge_index1=torch.tensor(edge_index1, dtype=torch.long, device=device)
edge_index2=torch.tensor(edge_index2, dtype=torch.long, device=device)
if edge_attr1!=None:
edge_attr1=torch.tensor(edge_attr1, dtype=torch.long, device=device)
edge_attr2=torch.tensor(edge_attr2, dtype=torch.long, device=device)
data=[x1, x2, edge_index1, edge_index2, edge_attr1, edge_attr2]
prediction=model(data)
#batchloss=batchloss+criterion(prediction[0],prediction[1],label)
cossim=F.cosine_similarity(prediction[0],prediction[1])
batchloss=batchloss+criterion2(cossim,label)
batchloss.backward(retain_graph=True)
optimizer.step()
loss = batchloss.item()
totalloss+=loss
main_index = main_index + len(batch)
loss=totalloss/main_index
epochs.set_description("Epoch (Loss=%g)" % round(loss,5))
vcount=0
for data, label in vset:
mode='epoch'+str(epoch)+'_'+str(vcount)
vcount+=1
print(label)
visualize(data, label,mode=mode)
'''for batch in trainloder:
batch=batch.to(device)
print(batch)
quit()
time_start=time.time()
model.forward(batch)
time_end=time.time()
print(time_end-time_start)
quit()'''