-
Notifications
You must be signed in to change notification settings - Fork 0
/
time_consumption.py
166 lines (158 loc) · 8.27 KB
/
time_consumption.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
from BOfrom_scratch import *
import pdb
from tqdm import tqdm
if __name__ == '__main__':
PT = True
num_features = 20
bounds = [(0,1) for _ in range(num_features)]
if PT:
emsize = 512
encoder = encoders.Linear(num_features,emsize)
bptt = 2010
hps = {'noise': 1e-4, 'outputscale': 1., 'lengthscale': .6, 'fast_computations': (False,False,False)}
ys = priors.fast_gp.get_batch_first(100000,20,num_features, hyperparameters=hps)[1]
# num_border_list = [1000,10000]
num_borders = 1000
batch_fraction = 8
draw_flag = False
data_augment = True
lr = 0.0008
epochs = 625
root_dir = f'/home/ypq/TransformersCanDoBayesianInference/myresults/GPfitting_augmentTrue_{num_features}feature'
model = MyTransformerModel(encoder, num_borders, emsize, 4, 2*emsize, 6, 0.0,
y_encoder=encoders.Linear(1, emsize), input_normalization=False,
# pos_encoder=positional_encodings.NoPositionalEncoding(emsize, bptt*2),
pos_encoder=positional_encodings.NoPositionalEncoding(emsize, bptt*2),
decoder=None
)
model.criterion = bar_distribution.FullSupportBarDistribution(bar_distribution.get_bucket_limits(num_borders, ys=ys))
model_path = f'{root_dir}/numborder{num_borders}_lr{lr}_epoch{epochs}_GPfitting.pth'
checkpoint = torch.load(model_path)
model.load_state_dict({k.replace('module.',''):v for k,v in checkpoint.items()})
model.eval()
# time consumption in one iteration
# n_init_list = [50,100,200,400,800,1600,3200,6400,12800]
# # n_init_list = [3200]
# ac = 'EI'
# iter_step = 1
# repeat_num = 10
# out_root_path = "./numerical_results/time_comparison"
# out_path = f"{out_root_path}/time_per_it_feature{num_features}_repeat{repeat_num}.xlsx"
# results = {}
# results['n_init'] = n_init_list
# results['PT time/it mean'] = [0]*(len(n_init_list))
# results['PT time/it min'] = [np.inf]*(len(n_init_list))
# results['PT time/it max'] = [-np.inf]*(len(n_init_list))
# results['GP time/it mean'] = [0]*(len(n_init_list))
# results['GP time/it min'] = [np.inf]*(len(n_init_list))
# results['GP time/it max'] = [-np.inf]*(len(n_init_list))
# for i in range(len(n_init_list)):
# n_init = n_init_list[i]
# for n in tqdm(range(repeat_num)):
# PTBO = PTBayesianOptimization(function, bounds, model,n_init=n_init,ac=ac)
# GPBO = BayesianOptimization(function, bounds,n_init=n_init,ac=ac)
# t1 = time()
# PT_x_max = PTBO.optimize(n_iter=iter_step)
# t2 = time()
# print(f"{n_init} PT time/it: {t2-t1}s")
# t3 = time()
# GP_x_max = GPBO.optimize(n_iter=iter_step)
# t4 = time()
# print(f"{n_init} GP time/it: {t4-t3}s")
# results['PT time/it mean'][i] += (t2 - t1)
# results['GP time/it mean'][i] += (t4 - t3)
# results['PT time/it min'][i] = min(results['PT time/it min'][i], t2-t1)
# results['GP time/it min'][i] = min(results['GP time/it min'][i], t4-t3)
# results['PT time/it max'][i] = max(results['PT time/it max'][i], t2-t1)
# results['GP time/it max'][i] = max(results['GP time/it max'][i], t4-t3)
# results['PT time/it mean'][i] /= repeat_num
# results['GP time/it mean'][i] /= repeat_num
# df = pd.DataFrame(results)
# df.to_excel(out_path,index=False)
# simple regret: regret value versus time
# 不同维度(5,20,40)，不同初始点数量(800,1600)，不同函数(rastrigin,ackley)
n_init = 3200
ac = 'EI'
iter_step = 1
repeat_num = 1
time_step = 10 # 1s for 800 and 4s for 1600
total_time = 100 # s
time_length = total_time // time_step
func_type="unimodel"
out_root_path = "./numerical_results/time_comparison"
out_path = f"{out_root_path}/simple_regret_vstime_{func_type}_feature{num_features}_init{n_init}_time{total_time}_step{time_step}_repeat{repeat_num}.xlsx"
writer = pd.ExcelWriter(out_path)
functions = Function(func_type)
functions = functions()
max_values = {"quadratic":0,"exponential":np.e * num_features,"log":np.log(2) * num_features,"rosenbrock":0,"rastrigin":0,"ackley":0}
scale_factors = {"quadratic":0.25*num_features,"exponential":(np.e-1)*num_features,"log":np.log(2) * num_features,\
"rosenbrock":((num_features-1)*100+num_features//2),"rastrigin":20.25*num_features,"ackley":4.7}
for function_index in range(0,3):
function = list(functions.values())[function_index]
function_name = list(functions.keys())[function_index]
max_value = max_values[function_name]
scale_factor = scale_factors[function_name]
results = {}
results['time'] = [i for i in range(time_step,total_time+time_step,time_step)]
results['PT mean'] = [0]*(time_length)
results['PT min'] = [np.inf]*(time_length)
results['PT max'] = [-np.inf]*(time_length)
results['GP mean'] = [0]*(time_length)
results['GP min'] = [np.inf]*(time_length)
results['GP max'] = [-np.inf]*(time_length)
for n in tqdm(range(repeat_num)):
x = [np.array([np.random.uniform(bounds[i][0], bounds[i][1]) for i in range(num_features)]) for _ in range(n_init)]
y = [function(i) for i in x]
init_point = (x[:],y[:])
print(len(y))
PTBO = PTBayesianOptimization(function, bounds, model,n_init=n_init,ac=ac,init_point=init_point)
t1 = time()
t2 = time()
i = 0
iter_num = 0
while (t2-t1) < total_time:
PT_x_max = PTBO.optimize(n_iter=iter_step)
t2 = time()
iter_num += 1
if i < time_length and \
(t2-t1) >= results['time'][i]:
while i < time_length and \
(t2-t1) >= results['time'][i]:
i += 1
results['PT mean'][i-1] += (max_value-function(PT_x_max)[0])/scale_factor
results['PT min'][i-1] = min(results['PT min'][i-1], (max_value-function(PT_x_max)[0])/scale_factor)
results['PT max'][i-1] = max(results['PT max'][i-1], (max_value-function(PT_x_max)[0])/scale_factor)
print(f"{function_name} PT: {t2-t1}s; regret value: {(max_value-function(PT_x_max)[0])/scale_factor}")
print(f"iter_num:{iter_num}")
init_point = (x[:],y[:])
print(len(y))
GPBO = BayesianOptimization(function, bounds,n_init=n_init,ac=ac,init_point=init_point)
t3 = time()
t4 = time()
i = 0
iter_num = 0
while (t4-t3) < total_time:
GP_x_max = GPBO.optimize(n_iter=iter_step)
t4 = time()
iter_num += 1
if i < time_length and \
(t4-t3) >= results['time'][i]:
while i < time_length and \
(t4-t3) >= results['time'][i]:
i += 1
results['GP mean'][i-1] += (max_value-function(GP_x_max)[0])/scale_factor
results['GP min'][i-1] = min(results['GP min'][i-1], (max_value-function(GP_x_max)[0])/scale_factor)
results['GP max'][i-1] = max(results['GP max'][i-1], (max_value-function(GP_x_max)[0])/scale_factor)
print(f"{function_name} GP : {t4-t3}s; regret value: {(max_value-function(GP_x_max)[0])/scale_factor}")
print(f"iter_num:{iter_num}")
tmp1 = [results['PT mean'][i] / repeat_num for i in range(total_time//time_step)]
tmp2 = [results['GP mean'][i] / repeat_num for i in range(total_time//time_step)]
results['PT mean'] = tmp1
results['GP mean'] = tmp2
df = pd.DataFrame(results)
df = df.replace(0,np.nan)
df = df.fillna(method='ffill')
df_name = pd.DataFrame({function_name:[]})
df_name.to_excel(writer,index=False,startrow=(function_index)*(total_time//time_step + 3))
df.to_excel(writer,index=False,startrow=(function_index)*(total_time//time_step + 3)+1)
writer.save()