-
Notifications
You must be signed in to change notification settings - Fork 81
/
Copy pathget_results.py
199 lines (188 loc) · 7.28 KB
/
get_results.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
189
190
191
192
193
194
195
196
197
198
199
import sys
import glob
import numpy as np
import matplotlib.pyplot as plt
def read_file(file_name):
file=open(file_name)
contents=file.readlines()
train_result_dict={}
val_result_dict={}
for line in contents:
if 'Epoch' not in line:
continue
epoch_str=line[line.find('Epoch')+5:]
batch_str=line[line.find('Batch')+5:]
epoch=int(epoch_str[epoch_str.find('[')+1:epoch_str.find(']')])+1
if 'Validation' in line:
if not val_result_dict.has_key(epoch):
val_result_dict[epoch]={}
val_result_dict[epoch]['loss']=[]
val_result_dict[epoch]['top1']=[]
val_result_dict[epoch]['top5']=[]
key=line[line.find('Validation'):]
value=key[key.find('=')+1:]
if 'Validation-cro' in key:
val_result_dict[epoch]['loss'].append(float(value))
elif 'Validation-acc' in key:
val_result_dict[epoch]['top1'].append(float(value))
elif 'Validation-top' in key:
val_result_dict[epoch]['top5'].append(float(value))
elif '[750]' in batch_str:
if not val_result_dict.has_key(epoch):
train_result_dict[epoch]={}
train_result_dict[epoch]['loss']=[]
train_result_dict[epoch]['top1']=[]
train_result_dict[epoch]['top5']=[]
loss=line[line.find('loss')+4:]
loss=loss[loss.find('(')+1:loss.find(')')]
top1=line[line.find('top1')+4:]
top1=top1[top1.find('(')+1:top1.find(')')]
top5=line[line.find('top5')+4:]
top5=top5[top5.find('(')+1:top5.find(')')]
try:
loss=float(loss)
top1=float(top1)
top5=float(top5)
except:
loss=loss if isinstance(loss,float) else -1
top1=top1 if isinstance(top1,float) else -1
top5=top5 if isinstance(top5,float) else -1
scale= 0.01 if top1>1 else 1
train_result_dict[epoch]['loss'].append(loss)
train_result_dict[epoch]['top1'].append(scale*top1)
train_result_dict[epoch]['top5'].append(scale*top5)
# elif array[5]=='arguments':
#print array[6:]
return train_result_dict,val_result_dict
def print_result(result_dict,style='acc'):
top1=[]
top5=[]
loss=[]
bias=0.0
scale=100.0
best=max
if style=='err':
bias=100.0
scale=-100.0
best=min
for epoch in range(1,len(result_dict)+1):
res=result_dict[epoch]
top1.append(bias+scale*res['top1'][-1])
if 'top5' in res.keys() and len(res['top5'])>0:
top5.append(bias+scale*res['top5'][-1])
loss.append(res['loss'][-1])
return top1,top5,loss,best
def sort_results(t_dict_list,test_idx=-1):
t_res=[]
t_list=[]
loss_list=[]
for t_dict in t_dict_list:
t_top1,_,t_loss,_=print_result(t_dict,style='acc')
t_list.append(t_top1)
loss_list.append(t_loss)
t_res.append(t_top1[-1])
if test_idx==-1:
mid=int( (len(t_res)-1)/2 )
index=np.argsort(t_res)[mid]
else:
index=test_idx
t_str='%6.3f%% (%6.3f +/- %6.3f, %6.3f)'%(t_res[index],np.mean(t_res),np.std(t_res), np.max(t_res))
top_curve=t_list[index]
loss_curve=loss_list[index]
return index,t_str,top_curve, loss_curve
def net_results(exp_name):
file_names= glob.glob('%s*.txt'%(exp_name))
train_list=[]
test_list=[]
valid_names=[]
for file_name in file_names:
train_dict,test_dict=read_file(file_name)
if len(train_dict)%40!=0:
continue
valid_names.append(file_name)
train_list.append(train_dict)
test_list.append(test_dict)
# print '\n',len(valid_names),valid_names
if len(test_list)<1:
return None,None,None,0
index,test_str,test_top1,test_loss=sort_results(test_list)
_,train_str,train_top1,train_loss=sort_results(train_list,index)
train_tuple=(train_str,train_top1,train_loss)
test_tuple=(test_str,test_top1,test_loss)
return train_tuple,test_tuple,valid_names[index],len(test_list)
def read_result(result_filename):
with open(result_filename,'r') as file:
contents=file.readlines()
print contents[0][:-1]
def process_exps(snapshot_dir, exp_names,override=False):
for exp_name in exp_names:
#original file
result_filename=snapshot_dir+'_result/'+exp_name[exp_name.rfind('/')+1:]+'.txt'
import os
if not override and os.path.isfile(result_filename):
read_result(result_filename)
continue #it has been processed
#read results
train_tuple,test_tuple,file_name,run_num=net_results(exp_name)
if run_num<1:
continue
train_str,train_top1,train_loss=train_tuple
test_str,test_top1,test_loss=test_tuple
result_str='%13s'%(exp_name[exp_name.rfind('\\')+1:])
result_str+=', testing: '+test_str
result_str+=', training: '+train_str
print 'run_num:%d %s'%(run_num,result_str)
#write to result file
dirname=result_filename[:result_filename.rfind('\\')]
if not os.path.exists(dirname):
os.makedirs(dirname)
origin_file=open(file_name)
contents=origin_file.readlines()
train_result_dict={}
val_result_dict={}
result_file=open(result_filename,'w')
result_file.write(result_str+'\n\n')
for oneline in contents[:10]:
result_file.write(oneline[32:])
if 'Start' in oneline:
break
#write to result file
for idx in range(len(train_top1)):
line='Epoch[%03d] %8s loss: %8.6f, acc: %6.3f; '%(idx+1,'training',train_loss[idx],train_top1[idx])
line+='%8s loss: %05.3f, acc: %6.3f\n'%('testing',test_loss[idx],test_top1[idx])
result_file.write(line)
result_file.close()
def get_exps(snapshot_dir,network_name):
txt_files=glob.glob('%s/%s/*.txt'%(snapshot_dir,network_name))
exp_names={}
for file_name in txt_files:
key=file_name[:file_name.rfind('_')]
depth=0
if key not in exp_names.keys():
try:
depth=int(key[key.find('_d')+2:key.find('w')])
except:
depth+=1
finally:
exp_names[key]=depth
import operator
sorted_x = sorted(exp_names.items(), key=operator.itemgetter(1))
return [name for name,depth in sorted_x]
def main():
from optparse import OptionParser
parser = OptionParser()
parser.add_option("--log-dir", type=str, default='./logfiles/cifar10',
help="directory of the log files, e.g. '--log-dir=./logfiles/cifar10'")
parser.add_option("-f", "--force",
action="store_true", dest="override", default=False,
help="override existing files")
args,_ = parser.parse_args()
snapshot_dir=args.log_dir
networks=['plain','side','fuse3','fuse6','resnet','half','cross']
for network in networks:
print 'network:'+network
exp_names=get_exps(snapshot_dir,network)
process_exps(snapshot_dir, exp_names,override=args.override)
print '\n'
if __name__ == '__main__':
main()