-
Notifications
You must be signed in to change notification settings - Fork 2
/
Copy pathanalyze_surrogate_shape_x_c_functions.py
214 lines (211 loc) · 8.86 KB
/
analyze_surrogate_shape_x_c_functions.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
import sys
sys.path.append("./mesh")
import numpy as np
import os
import pandas as pd
from PolyhedronMesh import PolyhedronMesh
from train_val_test_split_x_c_new1 import train_val_test_split
#%%
def compare(filelist_true, filstlist_pred, stress):
MAPE_list=[]
APE_list=[]
mrse_mean_list=[]
mrse_min_list=[]
mrse_max_list=[]
for file_true, file_pred in zip(filelist_true, filstlist_pred):
#mesh_p0=PolyhedronMesh()
#file_p0=file_true.replace('i90', 'i0')
#mesh_p0.load_from_torch(file_p0)
mesh_true=PolyhedronMesh()
mesh_true.load_from_torch(file_true)
mesh_pred=PolyhedronMesh()
mesh_pred.load_from_torch(file_pred)
#---------------------------------------
#X_true=mesh_p0.node
x_true=mesh_true.node
x_pred=mesh_pred.node
#disp_mean=((X_true-x_true)**2).sum(dim=1).sqrt().mean()
mrse=((x_pred-x_true)**2).sum(dim=1).sqrt()#/disp_mean
mrse_mean_list.append(mrse.mean().item())
mrse_max_list.append(mrse.max().item())
mrse_min_list.append(mrse.min().item())
#----------------------------------------
if stress == "VM":
s_true=mesh_true.element_data['VM']
s_pred=mesh_pred.element_data['VM']
elif stress == "S":
s_true=mesh_true.element_data['S']
s_pred=mesh_pred.element_data['S']
if stress == "VM":
#print(s_true.shape, s_pred.shape)
s_true_abs_mean=s_true.abs().mean().item()
s_true_abs_max=s_true.abs().max().item()
s_pred_abs_max=s_pred.abs().max().item()
PE=(s_pred-s_true).abs()/s_true_abs_mean
MAPE=PE.mean().item()
MAPE_list.append(MAPE)
APE=abs(s_true_abs_max-s_pred_abs_max)/s_true_abs_mean
APE_list.append(APE)
elif stress == "S":
s_true=s_true.view(s_true.shape[0],-1)
s_pred=s_pred.view(s_pred.shape[0],-1)
s_true_abs_mean=(s_true**2).sum(dim=1).sqrt().mean().item()
s_true_abs_max=(s_true**2).sum(dim=1).sqrt().max().item()
s_pred_abs_max=(s_pred**2).sum(dim=1).sqrt().max().item()
s_error=((s_true-s_pred)**2).sum(dim=1).sqrt()
PE=s_error/s_true_abs_mean
MAPE=PE.mean().item()
MAPE_list.append(MAPE)
APE=abs(s_true_abs_max-s_pred_abs_max)/s_true_abs_mean
APE_list.append(APE)
#----------------------------------------
mrse_mean_list=np.array(mrse_mean_list)
mrse_max_list=np.array(mrse_max_list)
mrse_min_list=np.array(mrse_min_list)
#----------------------------------------
MAPE_list=np.array(MAPE_list)
#if np.isnan(MAPE_list).sum()>0:
# print("np.isnan(MAPE_list).sum()>0: set nan to 1")
MAPE_list[np.isnan(MAPE_list)==True]=1
#----------------------------------------
APE_list=np.array(APE_list)
#if np.isnan(APE_list).sum()>0:
# print("np.isnan(APE_list).sum()>0: set nan to 1")
APE_list[np.isnan(APE_list)==True]=1
#if np.sum(APE_list>1)>0:
# print("np.sum(APE_list>1)>0")
return mrse_mean_list, mrse_max_list, mrse_min_list, MAPE_list, APE_list
#%%
def get_time_cost(filelist_true, filelist_pred):
time_true=[]
time_pred=[]
for file_true, file_pred in zip(filelist_true, filelist_pred):
mesh_true=PolyhedronMesh()
mesh_true.load_from_torch(file_true)
mesh_pred=PolyhedronMesh()
mesh_pred.load_from_torch(file_pred)
try:
time_pred.append(mesh_pred.mesh_data['time'][-1])
time_true.append(mesh_true.mesh_data['time'][-1])
if len(mesh_pred.mesh_data['time']) > 150:
print("time length > 150: OOD")
except:
pass
if len(time_pred) > 0:
time_true=np.array(time_true)
time_pred=np.array(time_pred)
time_cost=time_pred/(1e-10+time_true)
else:
time_cost=np.zeros(len(filelist_true))
return time_cost
#%%
def get_filelist(net, folder_data, folder_result, train_percent, test_or_val, refine):
(filelist_train, filelist_val, filelist_test,
shape_idlist_train, shape_idlist_val, shape_idlist_test)=train_val_test_split(folder_data, train_percent)
if test_or_val == 'test':
filelist_true=filelist_test
folder_pred=folder_result+net+"/"+str(train_percent)+"/matMean/test/"
elif test_or_val == 'val':
filelist_true=filelist_val
folder_pred=folder_result+net+"/"+str(train_percent)+"/matMean/val/"
else:
filelist_true=filelist_train
folder_pred=folder_result+net+"/"+str(train_percent)+"/matMean/train/"
filelist_pred=[]
for n in range(0, len(filelist_true)):
name_true=filelist_true[n]
name_true=name_true.replace('\\', '/')
name_true=name_true.split('/')[-1].split('.')[0]
name_true=name_true.replace('p20_i90','p18')
if refine == False:
name_pred=folder_pred+name_true+"_pred.pt"
else:
name_pred=folder_pred+name_true+"_pred_refine_R1.pt"
filelist_pred.append(name_pred)
idlist=[]
for n in range(0, len(filelist_pred)):
a=os.path.isfile(filelist_pred[n])
if a == True:
idlist.append(n)
error_counter=len(filelist_pred)-len(idlist)
filelist_true=np.array(filelist_true)
filelist_true=filelist_true[idlist]
filelist_pred=np.array(filelist_pred)
filelist_pred=filelist_pred[idlist]
return filelist_true, filelist_pred, error_counter
#%%
def get_table(net, folder_data, folder_result, train_percent=0.5, test_or_val='test', refine=False, stress='VM',
return_error_counter=False):
mrse_table=[]
MAPE_table=[]
APE_table=[]
time_table=[]
filelist_true, filelist_pred, error_counter=get_filelist(net, folder_data, folder_result,
train_percent, test_or_val, refine)
mrse_mean_list, mrse_max_list, mrse_min_list, MAPE_list, APE_list=compare(filelist_true, filelist_pred, stress)
mrse_mean=np.mean(mrse_mean_list)
#mrse_mean=np.median(mrse_mean_list)
#mrse_max=np.max(mrse_max_list)
mrse_max=np.max(mrse_mean_list)#paper
#mrse_min=np.min(mrse_min_list)
mrse_min=np.min(mrse_mean_list)
mrse=(mrse_mean, mrse_max, mrse_min)
MAPE_mean=np.mean(MAPE_list)
#MAPE_mean=np.median(MAPE_list)
MAPE_max=np.max(MAPE_list)
MAPE_min=np.min(MAPE_list)
MAPE=(MAPE_mean, MAPE_max, MAPE_min)
APE_mean=np.mean(APE_list)
#APE_mean=np.median(APE_list)
APE_max=np.max(APE_list)
APE_min=np.min(APE_list)
APE=(APE_mean, APE_max, APE_min)
time_cost=get_time_cost(filelist_true, filelist_pred)
time_cost_mean=np.mean(time_cost)
time_cost_max=np.max(time_cost)
time_cost_min=np.min(time_cost)
time_cost=(time_cost_mean, time_cost_max, time_cost_min)
mrse_table.append(mrse)
MAPE_table.append(MAPE)
APE_table.append(APE)
time_table.append(time_cost)
if return_error_counter == False:
return mrse_table, MAPE_table, APE_table, time_table
else:
return mrse_table, MAPE_table, APE_table, time_table, error_counter
#%%
def get_result(net, folder_data, folder_result, train_percent=0.5, test_or_val='test', stress='VM', refine=False):
filelist_true, filelist_pred, error=get_filelist(net, folder_data, folder_result,
train_percent, test_or_val, refine)
mrse_mean, mrse_max, mrse_min, MAPE_list, APE_list=compare(filelist_true, filelist_pred, stress)
time_cost=get_time_cost(filelist_true, filelist_pred)
return mrse_mean, mrse_max, MAPE_list, APE_list, time_cost, filelist_true, filelist_pred
#%%
def get_data_frame(net, name, folder_data, folder_result, test_or_val, stress, refine):
mrse, MAPE, APE, t, error=get_table(net, folder_data, folder_result, test_or_val=test_or_val,
stress=stress, refine=refine, return_error_counter=True)
frame=[]
frame.append('{a:.4f}'.format(a=mrse[0][0]))
frame.append('{a:.4f}'.format(a=mrse[0][1]))
frame.append('{a:.4f}%'.format(a=100*MAPE[0][0]))
if MAPE[0][0] > 1:
frame[2]='>100%'
frame.append('{a:.4f}%'.format(a=100*MAPE[0][1]))
if MAPE[0][1] > 1:
frame[3]='>100%'
frame.append('{a:.4f}%'.format(a=100*APE[0][0]))
if APE[0][0] > 1:
frame[4]='>100%'
frame.append('{a:.4f}%'.format(a=100*APE[0][1]))
if APE[0][1] > 1:
frame[5]='>100%'
if refine == True:
frame.append('{a:.4f}%'.format(a=100*t[0][0]))
frame.append('{a:.4f}%'.format(a=100*t[0][1]))
frame.append(error)
columns=['MRSE_avg', 'MRSE_max', 'MAPE_avg', 'MAPE_max', 'APE_avg', 'APE_max',
'Time_avg', 'Time_max', 'error']
else:
columns=['MRSE_avg', 'MRSE_max', 'MAPE_avg', 'MAPE_max', 'APE_avg', 'APE_max']
frame=pd.DataFrame([frame], columns=columns, index=[name])
return frame