-
Notifications
You must be signed in to change notification settings - Fork 2
/
gen_settings_from_base.py
125 lines (111 loc) · 5.76 KB
/
gen_settings_from_base.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
import pandas as pd
import os, sys
'''
E.g.,
python gen_settings_from_base.py traffic deepar deepar_cedar deepar_cedar 64
python gen_settings_from_base.py traffic deepar deepar_mmd deepar_mmd 64
'''
try:
dataset_name = sys.argv[1]
base_model_file = sys.argv[2]
target_model = sys.argv[3]
outfile_name = sys.argv[4]
new_batch_size = sys.argv[5]
except:
print ('Usage: python train.py <dataset_name> <base_model_file `deepar/wavenet`> <target_model `deepar_cedar`> <outfile_name `deepar_cedar`> <new_batch_size `64`>')
exit()
print(f'dataset_name:{dataset_name}, base_model_file:{base_model_file}, target_model:{target_model}, outfile_name:{outfile_name} new_batch_size:{new_batch_size}')
# read base model settings
base_model_file = f"experiments/{dataset_name}/{base_model_file}.csv"
print('base_model_file',base_model_file)
target_model_file = f"experiments/{dataset_name}/{outfile_name}.csv"
print('target_model_file',target_model_file)
assert os.path.exists(base_model_file), f"base model setting file not exist"
df = pd.read_csv(base_model_file,sep=';')
print(df)
best_param = df[df['seed']==2]
print('best_param', best_param)
base_model = best_param['algorithm'].values[0]
best_learning_rate = best_param['learning_rate'].values[0]
best_d_hidden = int(best_param['d_hidden'].values[0])
best_batch_size = int(best_param['batch_size'].values[0])
best_N = int(best_param['N'].values[0])
best_dropout = best_param['dropout'].values[0]
algorithm = target_model
seed = 0
if target_model.endswith('_mmd'):
gamma = [1.0, 1e-1, 1e-2, 1e-3, 1e-4, 1e-5, 1e-6, 1e-7, 1e-8]
loss_dis = ['']
elif target_model.endswith('_cedar'):
gamma = [1.0, 1e-1, 1e-2, 1e-3, 1e-4, 1e-5, 1e-6, 1e-7, 1e-8]
loss_dis = ['abs','sqaure']
reg = ['dd','ddd']
else:
print(f'{target_model} is not supported yet')
mmd = 'linear'
if base_model == 'deepar':
if target_model.endswith('_mmd'):
columns = "algorithm;seed;learning_rate;batch_size;d_hidden;N;mmd;gamma;dropout;score;in_progress;test_score;params_test;params_train"
params_test = "[test_set.d_lag, test_set.d_cov, d_emb, test_set.dim_output, d_hidden, dropout, N, mmd_type, gamma]"
params_train = "[training_set.d_lag, training_set.d_cov, d_emb, training_set.dim_output, d_hidden, dropout, N, mmd_type, gamma]"
else:
columns = "algorithm;seed;learning_rate;batch_size;d_hidden;N;mmd;loss_dis;reg;gamma;dropout;score;in_progress;test_score;params_test;params_train"
params_test = "[test_set.d_lag, test_set.d_cov, d_emb, test_set.dim_output, d_hidden, dropout, N, mmd_type, gamma, loss_dis, reg]"
params_train = "[training_set.d_lag, training_set.d_cov, d_emb, training_set.dim_output, d_hidden, dropout, N, mmd_type, gamma, loss_dis, reg]"
elif base_model == 'wavenet':
if target_model.endswith('_mmd'):
best_kernel_size = best_param['kernel_size'].values[0]
columns = "algorithm;seed;learning_rate;batch_size;d_hidden;kernel_size;N;mmd;gamma;dropout;score;in_progress;test_score;params_test;params_train"
params_test = "[test_set.d_lag, test_set.d_cov, d_emb, test_set.dim_output, d_hidden, N, kernel_size, mmd_type, gamma]"
params_train = "[training_set.d_lag, training_set.d_cov, d_emb, training_set.dim_output, d_hidden, N, kernel_size, mmd_type, gamma]"
else:
best_kernel_size = best_param['kernel_size'].values[0]
columns = "algorithm;seed;learning_rate;batch_size;d_hidden;kernel_size;N;mmd;loss_dis;reg;gamma;dropout;score;in_progress;test_score;params_test;params_train"
params_test = "[test_set.d_lag, test_set.d_cov, d_emb, test_set.dim_output, d_hidden, N, kernel_size, mmd_type, gamma, loss_dis, reg]"
params_train = "[training_set.d_lag, training_set.d_cov, d_emb, training_set.dim_output, d_hidden, N, kernel_size, mmd_type, gamma, loss_dis, reg]"
else:
print(f'{base_model} is not supported yet')
target_model_settings = pd.DataFrame(columns = columns.split(';'))
if target_model.endswith('_mmd'):
for gamma_v in gamma:
new_row = {}
new_row['algorithm'] = target_model
new_row['seed'] = 0
new_row['learning_rate'] = best_learning_rate
new_row['batch_size'] = new_batch_size
new_row['d_hidden'] = best_d_hidden
new_row['N'] = best_N
if base_model == 'wavenet':
new_row['kernel_size'] = best_kernel_size
new_row['mmd'] = mmd
new_row['gamma'] = gamma_v
new_row['dropout'] = best_dropout
new_row['in_progress'] = -1
new_row['params_test'] = params_test
new_row['params_train'] = params_train
target_model_settings = target_model_settings._append(new_row, ignore_index = True)
else:
for reg_v in reg:
for loss_dis_v in loss_dis:
for gamma_v in gamma:
new_row = {}
new_row['algorithm'] = target_model
new_row['seed'] = 0
new_row['learning_rate'] = best_learning_rate
new_row['batch_size'] = new_batch_size
new_row['d_hidden'] = best_d_hidden
new_row['N'] = best_N
if base_model == 'wavenet':
new_row['kernel_size'] = best_kernel_size
new_row['mmd'] = mmd
new_row['loss_dis'] = loss_dis_v
new_row['reg'] = reg_v
new_row['gamma'] = gamma_v
new_row['dropout'] = best_dropout
new_row['in_progress'] = -1
new_row['params_test'] = params_test
new_row['params_train'] = params_train
target_model_settings = target_model_settings._append(new_row, ignore_index = True)
target_model_settings.to_csv(target_model_file, sep=';', index=False)
print(target_model_settings)
print(f'{target_model_file} saved!')