|
1
|
|
|
# Author: Simon Blanke |
|
2
|
|
|
# Email: [email protected] |
|
3
|
|
|
# License: MIT License |
|
4
|
|
|
|
|
5
|
|
|
|
|
6
|
|
|
import random |
|
7
|
|
|
import numpy as np |
|
8
|
|
|
import multiprocessing |
|
9
|
|
|
|
|
10
|
|
|
from .checks import check_hyperactive_para, check_search_para |
|
11
|
|
|
|
|
12
|
|
|
|
|
13
|
|
|
def stop_warnings(): |
|
14
|
|
|
# because sklearn warnings are annoying when they appear 100 times |
|
15
|
|
|
def warn(*args, **kwargs): |
|
16
|
|
|
pass |
|
17
|
|
|
|
|
18
|
|
|
import warnings |
|
19
|
|
|
|
|
20
|
|
|
warnings.warn = warn |
|
21
|
|
|
|
|
22
|
|
|
|
|
23
|
|
|
class MainArgs: |
|
24
|
|
|
def __init__(self, X, y, memory, random_state, verbosity, warnings, ext_warnings): |
|
25
|
|
|
check_hyperactive_para(X, y, memory, random_state, verbosity) |
|
26
|
|
|
|
|
27
|
|
|
if not ext_warnings: |
|
28
|
|
|
stop_warnings() |
|
29
|
|
|
|
|
30
|
|
|
self._verb_ = None |
|
31
|
|
|
self.hyperactive_para = { |
|
32
|
|
|
"memory": memory, |
|
33
|
|
|
"random_state": random_state, |
|
34
|
|
|
"verbosity": verbosity, |
|
35
|
|
|
} |
|
36
|
|
|
|
|
37
|
|
|
self.X = X |
|
38
|
|
|
self.y = y |
|
39
|
|
|
self.verbosity = verbosity |
|
40
|
|
|
self.random_state = random_state |
|
41
|
|
|
self.memory = memory |
|
42
|
|
|
|
|
43
|
|
|
self.opt_para = dict() |
|
44
|
|
|
|
|
45
|
|
|
def search_args( |
|
46
|
|
|
self, search_config, max_time, n_iter, optimizer, n_jobs, scheduler, init_config |
|
47
|
|
|
): |
|
48
|
|
|
check_search_para( |
|
49
|
|
|
search_config, max_time, n_iter, optimizer, n_jobs, scheduler, init_config |
|
50
|
|
|
) |
|
51
|
|
|
|
|
52
|
|
|
self.search_para = { |
|
53
|
|
|
"search_config": search_config, |
|
54
|
|
|
"max_time": max_time, |
|
55
|
|
|
"n_iter": n_iter, |
|
56
|
|
|
"optimizer": optimizer, |
|
57
|
|
|
"n_jobs": n_jobs, |
|
58
|
|
|
"scheduler": scheduler, |
|
59
|
|
|
"init_config": init_config, |
|
60
|
|
|
} |
|
61
|
|
|
|
|
62
|
|
|
self.search_config = search_config |
|
63
|
|
|
self.max_time = max_time |
|
64
|
|
|
self.n_iter = n_iter |
|
65
|
|
|
self.optimizer = optimizer |
|
66
|
|
|
self.n_jobs = n_jobs |
|
67
|
|
|
self.scheduler = scheduler |
|
68
|
|
|
self.init_config = init_config |
|
69
|
|
|
|
|
70
|
|
|
self.model_list = list(self.search_config.keys()) |
|
71
|
|
|
self.n_models = len(self.model_list) |
|
72
|
|
|
|
|
73
|
|
|
if self.max_time: |
|
74
|
|
|
self.max_time = self.max_time * 3600 |
|
75
|
|
|
|
|
76
|
|
|
self.set_n_jobs() |
|
77
|
|
|
|
|
78
|
|
|
self._n_process_range = range(0, int(self.n_jobs)) |
|
79
|
|
|
|
|
80
|
|
|
if isinstance(optimizer, dict): |
|
81
|
|
|
self.optimizer = list(optimizer.keys())[0] |
|
82
|
|
|
self.opt_para = optimizer[self.optimizer] |
|
83
|
|
|
|
|
84
|
|
|
def _set_random_seed(self, thread=0): |
|
85
|
|
|
"""Sets the random seed separately for each thread (to avoid getting the same results in each thread)""" |
|
86
|
|
|
if self.n_jobs > 1 and not self.random_state: |
|
87
|
|
|
rand = np.random.randint(0, high=2 ** 32 - 2) |
|
88
|
|
|
random.seed(rand + thread) |
|
89
|
|
|
np.random.seed(rand + thread) |
|
90
|
|
|
|
|
91
|
|
|
elif self.random_state: |
|
92
|
|
|
rand = int(self.random_state) |
|
93
|
|
|
|
|
94
|
|
|
random.seed(rand + thread) |
|
95
|
|
|
np.random.seed(rand + thread) |
|
96
|
|
|
|
|
97
|
|
|
def set_n_jobs(self): |
|
98
|
|
|
"""Sets the number of jobs to run in parallel""" |
|
99
|
|
|
num_cores = multiprocessing.cpu_count() |
|
100
|
|
|
if self.n_jobs == -1 or self.n_jobs > num_cores: |
|
101
|
|
|
self.n_jobs = num_cores |
|
102
|
|
|
|
SimonBlanke /
Hyperactive
