|
1
|
|
|
# Author: Simon Blanke |
|
2
|
|
|
# Email: [email protected] |
|
3
|
|
|
# License: MIT License |
|
4
|
|
|
|
|
5
|
|
|
import os |
|
6
|
|
|
import multiprocessing |
|
7
|
|
|
from tqdm import tqdm |
|
8
|
|
|
|
|
9
|
|
|
from .optimizers import RandomSearchOptimizer |
|
10
|
|
|
from .run_search import run_search |
|
11
|
|
|
from .print_info import print_info |
|
12
|
|
|
|
|
13
|
|
|
from .hyperactive_results import HyperactiveResults |
|
14
|
|
|
|
|
15
|
|
|
|
|
16
|
|
|
class Hyperactive(HyperactiveResults): |
|
17
|
|
|
def __init__( |
|
18
|
|
|
self, |
|
19
|
|
|
verbosity=["progress_bar", "print_results", "print_times"], |
|
20
|
|
|
distribution={ |
|
21
|
|
|
"multiprocessing": { |
|
22
|
|
|
"initializer": tqdm.set_lock, |
|
23
|
|
|
"initargs": (tqdm.get_lock(),), |
|
24
|
|
|
} |
|
25
|
|
|
}, |
|
26
|
|
|
n_processes="auto", |
|
27
|
|
|
): |
|
28
|
|
|
super().__init__() |
|
29
|
|
|
if verbosity is False: |
|
30
|
|
|
verbosity = [] |
|
31
|
|
|
|
|
32
|
|
|
self.verbosity = verbosity |
|
33
|
|
|
self.distribution = distribution |
|
34
|
|
|
self.n_processes = n_processes |
|
35
|
|
|
|
|
36
|
|
|
self.search_ids = [] |
|
37
|
|
|
self.process_infos = {} |
|
38
|
|
|
|
|
39
|
|
|
self.progress_boards = {} |
|
40
|
|
|
|
|
41
|
|
|
def _add_search_processes( |
|
42
|
|
|
self, |
|
43
|
|
|
random_state, |
|
44
|
|
|
objective_function, |
|
45
|
|
|
optimizer, |
|
46
|
|
|
n_iter, |
|
47
|
|
|
n_jobs, |
|
48
|
|
|
max_score, |
|
49
|
|
|
memory, |
|
50
|
|
|
memory_warm_start, |
|
51
|
|
|
search_id, |
|
52
|
|
|
): |
|
53
|
|
|
if n_jobs == -1: |
|
54
|
|
|
n_jobs = multiprocessing.cpu_count() |
|
55
|
|
|
|
|
56
|
|
|
for _ in range(n_jobs): |
|
57
|
|
|
nth_process = len(self.process_infos) |
|
58
|
|
|
|
|
59
|
|
|
self.process_infos[nth_process] = { |
|
60
|
|
|
"random_state": random_state, |
|
61
|
|
|
"verbosity": self.verbosity, |
|
62
|
|
|
"nth_process": nth_process, |
|
63
|
|
|
"objective_function": objective_function, |
|
64
|
|
|
"optimizer": optimizer, |
|
65
|
|
|
"n_iter": n_iter, |
|
66
|
|
|
"max_score": max_score, |
|
67
|
|
|
"memory": memory, |
|
68
|
|
|
"memory_warm_start": memory_warm_start, |
|
69
|
|
|
"search_id": search_id, |
|
70
|
|
|
} |
|
71
|
|
|
|
|
72
|
|
|
def _default_opt(self, optimizer): |
|
73
|
|
|
if isinstance(optimizer, str): |
|
74
|
|
|
if optimizer == "default": |
|
75
|
|
|
optimizer = RandomSearchOptimizer() |
|
76
|
|
|
return optimizer |
|
77
|
|
|
|
|
78
|
|
|
def _default_search_id(self, search_id, objective_function): |
|
79
|
|
|
if not search_id: |
|
80
|
|
|
search_id = objective_function.__name__ |
|
81
|
|
|
return search_id |
|
82
|
|
|
|
|
83
|
|
|
def _init_progress_board(self, progress_board, search_id, search_space): |
|
84
|
|
|
data_c = None |
|
85
|
|
|
|
|
86
|
|
|
if progress_board: |
|
87
|
|
|
data_c = progress_board.init_paths(search_id, search_space) |
|
88
|
|
|
|
|
89
|
|
|
if progress_board.uuid not in self.progress_boards: |
|
90
|
|
|
self.progress_boards[progress_board.uuid] = progress_board |
|
91
|
|
|
|
|
92
|
|
|
return data_c |
|
93
|
|
|
|
|
94
|
|
|
@staticmethod |
|
95
|
|
|
def check_list(search_space): |
|
96
|
|
|
for key in search_space.keys(): |
|
97
|
|
|
search_dim = search_space[key] |
|
98
|
|
|
|
|
99
|
|
|
error_msg = ( |
|
100
|
|
|
"Value in '{}' of search space dictionary must be of type list".format( |
|
101
|
|
|
key |
|
102
|
|
|
) |
|
103
|
|
|
) |
|
104
|
|
|
|
|
105
|
|
|
if not isinstance(search_dim, list): |
|
106
|
|
|
print("Warning", error_msg) |
|
107
|
|
|
# raise ValueError(error_msg) |
|
108
|
|
|
|
|
109
|
|
|
def add_search( |
|
110
|
|
|
self, |
|
111
|
|
|
objective_function, |
|
112
|
|
|
search_space, |
|
113
|
|
|
n_iter, |
|
114
|
|
|
search_id=None, |
|
115
|
|
|
optimizer="default", |
|
116
|
|
|
n_jobs=1, |
|
117
|
|
|
initialize={"grid": 4, "random": 2, "vertices": 4}, |
|
118
|
|
|
max_score=None, |
|
119
|
|
|
random_state=None, |
|
120
|
|
|
memory=True, |
|
121
|
|
|
memory_warm_start=None, |
|
122
|
|
|
progress_board=None, |
|
123
|
|
|
): |
|
124
|
|
|
optimizer = self._default_opt(optimizer) |
|
125
|
|
|
search_id = self._default_search_id(search_id, objective_function) |
|
126
|
|
|
progress_collector = self._init_progress_board( |
|
127
|
|
|
progress_board, search_id, search_space |
|
128
|
|
|
) |
|
129
|
|
|
|
|
130
|
|
|
self.check_list(search_space) |
|
131
|
|
|
|
|
132
|
|
|
optimizer.init(search_space, initialize, progress_collector) |
|
133
|
|
|
|
|
134
|
|
|
self._add_search_processes( |
|
135
|
|
|
random_state, |
|
136
|
|
|
objective_function, |
|
137
|
|
|
optimizer, |
|
138
|
|
|
n_iter, |
|
139
|
|
|
n_jobs, |
|
140
|
|
|
max_score, |
|
141
|
|
|
memory, |
|
142
|
|
|
memory_warm_start, |
|
143
|
|
|
search_id, |
|
144
|
|
|
) |
|
145
|
|
|
|
|
146
|
|
|
def run(self, max_time=None, _test_st_backend=False): |
|
147
|
|
|
for nth_process in self.process_infos.keys(): |
|
148
|
|
|
self.process_infos[nth_process]["max_time"] = max_time |
|
149
|
|
|
|
|
150
|
|
|
# open progress board |
|
151
|
|
|
if not _test_st_backend: |
|
152
|
|
|
for progress_board in self.progress_boards.values(): |
|
153
|
|
|
progress_board.open_dashboard() |
|
154
|
|
|
|
|
155
|
|
|
self.results_list = run_search( |
|
156
|
|
|
self.process_infos, self.distribution, self.n_processes |
|
157
|
|
|
) |
|
158
|
|
|
|
|
159
|
|
|
for results in self.results_list: |
|
160
|
|
|
nth_process = results["nth_process"] |
|
161
|
|
|
|
|
162
|
|
|
print_info( |
|
163
|
|
|
verbosity=self.process_infos[nth_process]["verbosity"], |
|
164
|
|
|
objective_function=self.process_infos[nth_process][ |
|
165
|
|
|
"objective_function" |
|
166
|
|
|
], |
|
167
|
|
|
best_score=results["best_score"], |
|
168
|
|
|
best_para=results["best_para"], |
|
169
|
|
|
best_iter=results["best_iter"], |
|
170
|
|
|
eval_times=results["eval_times"], |
|
171
|
|
|
iter_times=results["iter_times"], |
|
172
|
|
|
n_iter=self.process_infos[nth_process]["n_iter"], |
|
173
|
|
|
) |
|
174
|
|
|
|
SimonBlanke /
Hyperactive
