-
Notifications
You must be signed in to change notification settings - Fork 0
/
utils.py
266 lines (206 loc) · 8.88 KB
/
utils.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
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
# -*- coding: utf-8 -*-
from settings.config_file import *
import os
import matplotlib.pyplot as plt
from sklearn.metrics import roc_auc_score, average_precision_score, auc, precision_recall_curve,matthews_corrcoef, balanced_accuracy_score,accuracy_score
def set_seed(num_seed=num_seed):
# os.CUBLAS_WORKSPACE_CONFIG="4096:8"
# torch.backends.cudnn.deterministic = True
# torch.backends.cudnn.benchmark = False
torch.manual_seed(num_seed)
torch.cuda.manual_seed_all(num_seed)
np.random.seed(num_seed)
random.seed(num_seed)
def aucPerformance(y_true, y_pred):
y_true = y_true.flatten().cpu()
y_pred = y_pred.flatten().cpu()
roc_auc = roc_auc_score(y_true, y_pred)
precision, recall, _ = precision_recall_curve(y_true, y_pred)
auc_pr = auc(recall, precision)
return roc_auc*100, auc_pr*100
def mcc_score(y_true, y_pred):
y_true = y_true.flatten().cpu()
y_pred = y_pred.flatten().cpu()
mcc = matthews_corrcoef(y_true, y_pred)
return mcc
def balanced_acc_score(y_true, y_pred):
y_true = y_true.flatten().cpu()
y_pred = y_pred.flatten().cpu()
balanced_acc = balanced_accuracy_score(y_true, y_pred)
return balanced_acc
def acc_score(y_true, y_pred):
y_true = y_true.flatten().cpu()
y_pred = y_pred.flatten().cpu()
acc_score = accuracy_score(y_true, y_pred)
return acc_score
def build_feature(function,option,num_function,e_search_space): # creer les feature d un graph provenant d un submodel
type_encoding= config["param"]["encoding_method"]
total_choices =int(config["param"]["total_choices"])
total_function =int(config["param"]["total_function"])
max_option =int(config["param"]["max_option"])
if type_encoding=="one_hot":
d= np.zeros((total_choices), dtype=int)
# print("option===",option)
d[option[1]]=1
elif type_encoding =="embedding":
d=option[2]
elif type_encoding =="index_embedding":
if config["param"]["feature_size_choice"] =="total_functions":
d= np.zeros((total_function), dtype=int)
d[num_function]= list(e_search_space[function]).index(option[0])+1
else:
d= np.zeros((total_choices), dtype=int)
d[option[1]]=list(e_search_space[function]).index(option[0])+1
elif type_encoding=="OneHot":
d= np.zeros((max_option), dtype=int)
pos=2
return d
def get_nodes_features(model_config,e_search_space):
nodes_features_list=[]
model_config_choices=[]
num_function=0
for function,option in model_config.items():
# print("function===",function)
feat = build_feature(function,option,num_function,e_search_space)
num_function+=1
nodes_features_list.append(feat)
model_config_choices.append((function,option[1])) # (AV ANT) quand j enleve la conversion en str cela provoque plus tard des erreurs incomprises au niveau du batching
x=np.array(nodes_features_list)
x = torch.tensor(x,dtype=torch.float32)
return x
def get_edge_index(model_config):
edge_dict={}
node_idx={}
# variable to controll the random sampling to make sure options are distribute uniformly
idx=0
for functions,options in model_config.items():
node_idx[functions]=idx
idx+=1
edge_dict['gnnConv1']=["normalize1",'dropout1','activation1']
edge_dict['aggregation1']=["gnnConv1"]
edge_dict['multi_head1']=["gnnConv1"]
edge_dict['hidden_channels1']= ["gnnConv1"]
edge_dict['normalize1']=["dropout1",'activation1']
edge_dict['dropout1']=["activation1"]
edge_dict['activation1']=["gnnConv2"]
edge_dict['gnnConv2']= ["normalize2",'dropout2','activation2']
edge_dict['aggregation2']=["gnnConv2"]
edge_dict['multi_head2']=["gnnConv2"]
edge_dict['hidden_channels2']= ["gnnConv2"]
edge_dict['normalize2']=["dropout2",'activation2']
edge_dict['dropout2']=["activation2"]
if config["dataset"]['type_task']=="graph_classification":
edge_dict['activation2']= ["pooling"]
edge_dict['pooling']=['criterion']
else:
edge_dict['activation2']= ["criterion"]
edge_dict['lr']= ["criterion","weight_decay"]
edge_dict['weight_decay']=["criterion","lr"]
edge_dict["criterion"]=["optimizer"]
edge_dict["optimizer"]=[]
source=[]
target=[]
edge_index=[]
for function,options in model_config.items():
# source.append(node_idx[function])
# target.append(node_idx[function])
if config['param']['type_input_graph']=="undirected":
for function2,options2 in model_config.items():
source.append(node_idx[function])
target.append(node_idx[function2])
# a=node_idx[function]
# b=node_idx[function2]
# print(f"Edge between {a} and {b}")
else:
for elt in edge_dict[function]:
source.append(node_idx[function])
target.append(node_idx[elt])
# a=node_idx[function]
# b=node_idx[elt]
# print(f"Edge between {a} and {b}")
edge_index.append(source)
edge_index.append(target)
edge_index=np.array(edge_index)
edge_index=torch.tensor(edge_index,dtype=torch.long)
return edge_index
def manage_budget():
budget =int(config["param"]["budget"])
k= int(config["param"]["k"])
z_sample= int(config["param"]["z_sample"])
z_topk= int(config["param"]["z_topk"])
z_final= int(config["param"]["z_final"])
n= int((budget-(k*z_topk)-z_final)/z_sample)
if n<=0:
print("Configuration error, Please change budget realated parameters")
raise SystemExit
else:
add_config("param","n",n)
def seed_worker(worker_id):
worker_seed = torch.initial_seed() % 2**32
numpy.random.seed(worker_seed)
random.seed(worker_seed)
g = torch.Generator()
g.manual_seed(0)
return g
def load_data(train_dataset,batch_size,num_workers,worker_init_fn,generator):
loader = DataLoader(
train_dataset,
batch_size=batch_size,
num_workers=num_workers,
worker_init_fn=seed_worker,
generator=generator)
return loader
def Generate_time_cost():
dataset_construction = float(config["time"]["distribution_time"])
predictor_training = float(config["time"]["predictor_training_time"])
gnn_encoding = float(config["time"]["sampling_time"])
topk_gnn_prediction = float(config["time"]["pred_time"])
topk_training = float(config["time"]["best_acc_time"])
total=float(config["time"]["total_search_time"])
# Make a random dataset:
height = [dataset_construction,predictor_training,gnn_encoding+topk_gnn_prediction,topk_training]
bars = ('Predictor training dataset construction','predictor training', 'top-k gnn prediction', 'top k gnn training')
y_pos = np.arange(len(bars))
fig,ax = plt.subplots()
# Create bars
plt.barh(y_pos, height)
plt.title("Running time details on {dataset_name} dataset")
plt.ylabel("running time(seconds)")
# Create names on the x-axis
plt.xticks(y_pos, bars)
plt.grid()
plt.show()
fig.savefig(f'{config["path"]["plots_folder"]}/{dataset_name}_timeCost_details_bar.pdf',bbox_inches="tight")
# explosion
fig, ax = plt.subplots(figsize=(25,10), subplot_kw=dict(aspect="equal"))
# Pie Chart
plt.pie(height, labels=bars,
autopct='%1.1f%%', pctdistance=0.85)
# draw circle
centre_circle = plt.Circle((0, 0), 0.60, fc='white')
fig = plt.gcf()
# Adding Circle in Pie chart
fig.gca().add_artist(centre_circle)
# Adding Title of chart
plt.title("Running time details on {dataset_name} dataset")
# Displaing Chart
plt.show()
fig.savefig(f'{config["path"]["plots_folder"]}/{dataset_name}_timeCost_details_pie.pdf',bbox_inches="tight")
bars = ('GraphNAS','RS', 'GAS', 'Auto-GNAS',"GraphNAP")
if config["dataset"]["dataset_name"]== "Cora":
height = [12960,12240,11520,3240,total]
elif config["dataset"]["dataset_name"]== "Citeseer":
height = [13320,13248,13680,4140,total]
elif config["dataset"]["dataset_name"]== "Pubmed":
height = [18360,18360,16560,5760,total]
y_pos = np.arange(len(bars))
fig,ax = plt.subplots()
# Create bars
plt.bar(y_pos, height)
plt.title("Running time Comparison on {dataset_name} dataset")
plt.ylabel("running time(seconds)")
# Create names on the x-axis
plt.xticks(y_pos, bars)
plt.grid()
plt.show()
fig.savefig(f'{config["path"]["plots_folder"]}/{dataset_name}_timeCost_comparison.pdf',bbox_inches="tight")