-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathanalyze_logs.py
250 lines (221 loc) · 13.7 KB
/
analyze_logs.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
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
import argparse
import os
import re
from itertools import groupby
import scienceplots
import matplotlib.pyplot as plt
import numpy as np
import seaborn as sns
from mmpretrain.utils import load_json_log
def get_log_dicts(filename):
json_logs = filename
for json_log in json_logs:
assert json_log.endswith('.json')
log_dicts = [load_json_log(json_log) for json_log in json_logs]
return log_dicts
# def plot_curve(log_dicts, args):
# """Plot train metric-iter graph."""
# # set style
# try:
# import seaborn as sns
# sns.set_style(args.style)
# except ImportError:
# pass
# # set plot window size
# wind_w, wind_h = args.window_size.split('*')
# wind_w, wind_h = int(wind_w), int(wind_h)
# plt.figure(figsize=(wind_w, wind_h))
# # get legends and metrics
# legends = 'accu'
# metrics = args.keys
# # plot curves from log_dicts by metrics
# plot_curve_helper(log_dicts, metrics, args, legends)
# # set title and show or save
# if args.title is not None:
# plt.title(args.title)
# if args.out is None:
# plt.show()
# else:
# print(f'save curve to: {args.out}')
# plt.savefig(args.out)
# plt.cla()
def plot_loss(log_dicts):
log_train=log_dicts['train']
loss=[log['loss'] for log in log_train]
iterations=[log['step'] for log in log_train]
plt.plot(iterations,loss)
plt.xlabel('iterations')
plt.ylabel('loss')
plt.show()
def plot_accu(log_dicts):
log_train=log_dicts['val']
accu=[log['accuracy/top1'] for log in log_train]
iterations=[50*i for i,_ in enumerate(log_train)]
plt.plot(iterations,accu)
plt.xlabel('iterations')
plt.ylabel('accu')
plt.show()
def get_eps(filename):
eps=[]
for fname in filename:
temp1=fname.index('eps')+3
temp2=fname.index('/',temp1,temp1+10)
eps.append(float(fname[temp1:temp2]))
return eps
def plot_loss_curve(filename,dataset='cifar10',kind='train'):
log_dicts=get_log_dicts(filename)
eps=get_eps(filename)
if kind=='train':
y_label=['loss','loss']
elif kind=='val':
y_label=['accu','accuracy/top1']
for i in range(len(log_dicts)):
logs=log_dicts[i][kind]
if kind=='train':
iterations=[log['step'] for log in logs]
else:
iterations=[50*i for i,_ in enumerate(logs)]
y=[log[y_label[1]] for log in logs]
# sns.lineplot(iterations,y,label='eps={}'.format(eps[i]))
plt.plot(iterations,y,label='eps={}'.format(eps[i]))
plt.xlabel('iterations')
plt.ylabel(y_label[1])
plt.legend(loc=1)
plt.savefig(f'results/{dataset}/eps_to_{y_label[0]}.png')
# plt.show()
def plot_curve_defer(filename,filename_defer,dataset,kind='train'):
log_dicts=get_log_dicts(filename)
log_dicts_defer=get_log_dicts(filename_defer)
eps=get_eps(filename)
color_list=['b','g','r','c','m','y','k','w']
plt.style.use(['science','ieee'])
# plt.style.use('science')
s_=2.5
if kind=='train':
y_label=['loss','loss']
elif kind=='val':
y_label=['accu','accuracy/top1']
for i in range(len(log_dicts)):
logs=log_dicts[i][kind]
if kind=='train':
iterations=[log['step'] for log in logs]
else:
iterations=[20*i for i,_ in enumerate(logs)]
y=[log[y_label[1]] for log in logs]
# sns.lineplot(iterations,y,label='eps={}'.format(eps[i]))
plt.plot(iterations,y,label='$\delta$={}'.format(eps[i]),color=color_list[i],linewidth=s_)
for i in range(len(log_dicts_defer)):
logs_defer=log_dicts_defer[i][kind]
if kind=='train':
iterations=[log['step'] for log in logs_defer]
else:
iterations=[20*i for i,_ in enumerate(logs_defer)]
y_defer=[log[y_label[1]] for log in logs_defer]
# sns.lineplot(iterations,y,label='eps={}'.format(eps[i]))
plt.plot(iterations,y_defer,linestyle=':',color=color_list[i],linewidth=s_)
# plt.plot(iterations,y_defer,label='defer eps={}'.format(eps[i]),linestyle=':',color=color_list[i])
plt.tick_params(axis='both',labelsize=10)
plt.xlabel('iterations',size=20)
plt.ylabel(y_label[1],size=20)
# plt.title(dataset)
plt.legend(loc=1)
plt.savefig(f'results/{dataset}/{dataset}_compare_eps_to_{y_label[0]}.png')
plt.close()
if __name__=="__main__":
# filename=['work_dirs_20240104/vit_config_cifar10_bs64_p16_384_eps0/20231220_104047/vis_data/20231220_104047.json']
filename_cifar10=[
# 'work_dirs_20240104/vit_config_cifar10_bs64_p16_384_eps1e-2/20231221_001721/vis_data/20231221_001721.json',
'work_dirs_20240104/vit_config_cifar10_bs64_p16_384_eps5e-3/20231221_001731/vis_data/20231221_001731.json',
'work_dirs_20240104/vit_config_cifar10_bs64_p16_384_eps3e-3/20231221_091823/vis_data/20231221_091823.json',
# 'work_dirs_20240104/vit_config_cifar10_bs64_p16_384_eps25e-4/20231221_160556/vis_data/20231221_160556.json',
'work_dirs_20240104/vit_config_cifar10_bs64_p16_384_eps2e-3/20231221_091725/vis_data/20231221_091725.json',
'work_dirs_20240104/vit_config_cifar10_bs64_p16_384_eps1e-3/20231220_143917/vis_data/20231220_143917.json',
'work_dirs_20240104/vit_config_cifar10_bs64_p16_384_eps1e-4/20231220_143935/vis_data/20231220_143935.json',
'work_dirs_20240104/vit_config_cifar10_bs64_p16_384_eps0/20231220_104047/vis_data/20231220_104047.json',
]
filename_cifar10_defer=[
# 'work_dirs_20240104/vit_config_cifar10_bs64_p16_384_eps1e-2_defer/20231224_094114/vis_data/20231224_094114.json',
'work_dirs_20240104/vit_config_cifar10_bs64_p16_384_eps5e-3_defer/20231223_082410/vis_data/20231223_082410.json',
'work_dirs_20240104/vit_config_cifar10_bs64_p16_384_eps3e-3_defer/20231228_221811/vis_data/20231228_221811.json',
# 'work_dirs_20240104/vit_config_cifar10_bs64_p16_384_eps25e-4_defer/20231222_161738/vis_data/20231222_161738.json',
'work_dirs_20240104/vit_config_cifar10_bs64_p16_384_eps2e-3_defer/20231222_091434/vis_data/20231222_091434.json',
'work_dirs_20240104/vit_config_cifar10_bs64_p16_384_eps1e-3_defer/20231222_091625/vis_data/20231222_091625.json',
'work_dirs_20240104/vit_config_cifar10_bs64_p16_384_eps1e-4_defer/20231223_082205/vis_data/20231223_082205.json',
]
filename_cifar100=[
# 'work_dirs_20240104/vit_config_cifar100_bs64_p16_384_eps1e-2/20231228_185107/vis_data/20231228_185107.json',
'work_dirs_20240104/vit_config_cifar100_bs64_p16_384_eps5e-3/20231227_185319/vis_data/20231227_185319.json',
'work_dirs_20240104/vit_config_cifar100_bs64_p16_384_eps3e-3/20231225_213107/vis_data/20231225_213107.json',
# 'work_dirs_20240104/vit_config_cifar100_bs64_p16_384_eps25e-4/20231226_092031/vis_data/20231226_092031.json',
'work_dirs_20240104/vit_config_cifar100_bs64_p16_384_eps2e-3/20231225_213039/vis_data/20231225_213039.json',
'work_dirs_20240104/vit_config_cifar100_bs64_p16_384_eps1e-3/20231225_090815/vis_data/20231225_090815.json',
'work_dirs_20240104/vit_config_cifar100_bs64_p16_384_eps1e-4/20231225_090757/vis_data/20231225_090757.json',
'work_dirs_20240104/vit_config_cifar100_bs64_p16_384_eps0/20231224_094538/vis_data/20231224_094538.json',
]
filename_cifar100_defer=[
# 'work_dirs_20240104/vit_config_cifar100_bs64_p16_384_eps1e-2_defer/20231228_230423/vis_data/20231228_230423.json',
'work_dirs_20240104/vit_config_cifar100_bs64_p16_384_eps5e-3_defer/20231228_092519/vis_data/20231228_092519.json',
'work_dirs_20240104/vit_config_cifar100_bs64_p16_384_eps3e-3_defer/20231227_091425/vis_data/20231227_091425.json',
# 'work_dirs_20240104/vit_config_cifar100_bs64_p16_384_eps25e-4_defer/20231227_105922/vis_data/20231227_105922.json',
'work_dirs_20240104/vit_config_cifar100_bs64_p16_384_eps2e-3_defer/20231226_213722/vis_data/20231226_213722.json',
'work_dirs_20240104/vit_config_cifar100_bs64_p16_384_eps1e-3_defer/20231226_213651/vis_data/20231226_213651.json',
'work_dirs_20240104/vit_config_cifar100_bs64_p16_384_eps1e-4_defer/20231226_091558/vis_data/20231226_091558.json',
]
filename_cub=[
# 'work_dirs_20240104/vit_config_cub_bs64_p16_384_eps1e-2/20240101_002929/vis_data/20240101_002929.json',
'work_dirs_20240104/vit_config_cub_bs64_p16_384_eps5e-3/20231231_092736/vis_data/20231231_092736.json',
'work_dirs_20240104/vit_config_cub_bs64_p16_384_eps4e-3/20240103_224711/vis_data/20240103_224711.json',
'work_dirs_20240104/vit_config_cub_bs64_p16_384_eps3e-3/20231231_092653/vis_data/20231231_092653.json',
# 'work_dirs_20240104/vit_config_cub_bs64_p16_384_eps25e-4/20231230_081353/vis_data/20231230_081353.json',
'work_dirs_20240104/vit_config_cub_bs64_p16_384_eps2e-3/20231230_083004/vis_data/20231230_083004.json',
'work_dirs_20240104/vit_config_cub_bs64_p16_384_eps1e-3/20231229_090717/vis_data/20231229_090717.json',
'work_dirs_20240104/vit_config_cub_bs64_p16_384_eps1e-4/20231229_090654/vis_data/20231229_090654.json',
'work_dirs_20240104/vit_config_cub_bs64_p16_384_eps0/20231228_185803/vis_data/20231228_185803.json',
]
filename_cub_defer=[
# 'work_dirs_20240104/vit_config_cub_bs64_p16_384_eps1e-2_defer/20240101_024903/vis_data/20240101_024903.json',
'work_dirs_20240104/vit_config_cub_bs64_p16_384_eps5e-3_defer/20231231_205423/vis_data/20231231_205423.json',
'work_dirs_20240104/vit_config_cub_bs64_p16_384_eps4e-3_defer/20240108_172356/vis_data/20240108_172356.json',
'work_dirs_20240104/vit_config_cub_bs64_p16_384_eps3e-3_defer/20240102_090515/vis_data/20240102_090515.json',
# 'work_dirs_20240104/vit_config_cub_bs64_p16_384_eps25e-4_defer/20231230_162009/vis_data/20231230_162009.json',
'work_dirs_20240104/vit_config_cub_bs64_p16_384_eps2e-3_defer/20231230_165231/vis_data/20231230_165231.json',
'work_dirs_20240104/vit_config_cub_bs64_p16_384_eps1e-3_defer/20231229_180646/vis_data/20231229_180646.json',
'work_dirs_20240104/vit_config_cub_bs64_p16_384_eps1e-4_defer/20231229_192028/vis_data/20231229_192028.json',
]
filename_oxford_pets=[
'work_dirs_20240104/vit_config_oxford_iii_pets_bs64_p16_384_eps1e-2/20240103_224412/vis_data/20240103_224412.json',
'work_dirs_20240104/vit_config_oxford_iii_pets_bs64_p16_384_eps5e-3/20240103_141004/vis_data/20240103_141004.json',
# 'work_dirs_20240104/vit_config_oxford_iii_pets_bs64_p16_384_eps4e-3/20240104_024954/vis_data/20240104_024954.json',
'work_dirs_20240104/vit_config_oxford_iii_pets_bs64_p16_384_eps3e-3/20240102_213027/vis_data/20240102_213027.json',
# 'work_dirs_20240104/vit_config_cub_bs64_p16_384_eps25e-4/20231230_081353/vis_data/20231230_081353.json',
'work_dirs_20240104/vit_config_oxford_iii_pets_bs64_p16_384_eps2e-3/20240101_233511/vis_data/20240101_233511.json',
'work_dirs_20240104/vit_config_oxford_iii_pets_bs64_p16_384_eps1e-3/20240101_094637/vis_data/20240101_094637.json',
'work_dirs_20240104/vit_config_oxford_iii_pets_bs64_p16_384_eps1e-4/20240101_122210/vis_data/20240101_122210.json',
'work_dirs_20240104/vit_config_oxford_iii_pets_bs64_p16_384_eps0/20240101_094449/vis_data/20240101_094449.json',
]
filename_oxford_pets_defer=[
'work_dirs_20240104/vit_config_oxford_iii_pets_bs64_p16_384_eps1e-2_defer/20240103_141808/vis_data/20240103_141808.json',
'work_dirs_20240104/vit_config_oxford_iii_pets_bs64_p16_384_eps5e-3_defer/20240103_091741/vis_data/20240103_091741.json',
# 'work_dirs_20240104/vit_config_oxford_iii_pets_bs64_p16_384_eps4e-3_defer/20240104_074943/vis_data/20240104_074943.json',
'work_dirs_20240104/vit_config_oxford_iii_pets_bs64_p16_384_eps3e-3_defer/20240102_154523/vis_data/20240102_154523.json',
# 'work_dirs_20240104/vit_config_cub_bs64_p16_384_eps25e-4/20231230_081353/vis_data/20231230_081353.json',
'work_dirs_20240104/vit_config_oxford_iii_pets_bs64_p16_384_eps2e-3_defer/20240102_090215/vis_data/20240102_090215.json',
'work_dirs_20240104/vit_config_oxford_iii_pets_bs64_p16_384_eps1e-3_defer/20240101_163511/vis_data/20240101_163511.json',
'work_dirs_20240104/vit_config_oxford_iii_pets_bs64_p16_384_eps1e-4_defer/20240101_195412/vis_data/20240101_195412.json',
]
eps=get_eps(filename_cifar100)
print(eps)
# plot_curve_defer(filename_cifar10,filename_cifar10_defer,dataset='cifar10',kind='val')
# plot_curve_defer(filename_cifar10,filename_cifar10_defer,dataset='cifar10',kind='train')
# plot_curve_defer(filename_cifar100,filename_cifar100_defer,dataset='cifar100',kind='val')
# plot_curve_defer(filename_cifar100,filename_cifar100_defer,dataset='cifar100',kind='train')
plot_curve_defer(filename_cub,filename_cub_defer,dataset='cub',kind='val')
plot_curve_defer(filename_cub,filename_cub_defer,dataset='cub',kind='train')
# plot_curve_defer(filename_oxford_pets,filename_oxford_pets_defer,dataset='oxford_pets',kind='val')
# plot_curve_defer(filename_oxford_pets,filename_oxford_pets_defer,dataset='oxford_pets',kind='train')
# plot_loss_curve(filename_cifar100,dataset='cifar100',kind='val')
exit()
log_dicts=get_log_dicts(filename)[3]
# print(log_dicts)
plot_loss(log_dicts)
plot_accu(log_dicts)