|
1
|
|
|
# Author: Simon Blanke |
|
2
|
|
|
# Email: [email protected] |
|
3
|
|
|
# License: MIT License |
|
4
|
|
|
|
|
5
|
|
|
|
|
6
|
|
|
import multiprocessing as mp |
|
7
|
|
|
from tqdm import tqdm |
|
8
|
|
|
|
|
9
|
|
|
from .optimizers import RandomSearchOptimizer |
|
10
|
|
|
from .run_search import run_search |
|
11
|
|
|
|
|
12
|
|
|
from .results import Results |
|
13
|
|
|
from .print_results import PrintResults |
|
14
|
|
|
from .search_space import SearchSpace |
|
15
|
|
|
|
|
16
|
|
|
|
|
17
|
|
|
class Hyperactive: |
|
18
|
|
|
def __init__( |
|
19
|
|
|
self, |
|
20
|
|
|
verbosity=["progress_bar", "print_results", "print_times"], |
|
21
|
|
|
distribution="multiprocessing", |
|
22
|
|
|
n_processes="auto", |
|
23
|
|
|
): |
|
24
|
|
|
super().__init__() |
|
25
|
|
|
if verbosity is False: |
|
26
|
|
|
verbosity = [] |
|
27
|
|
|
|
|
28
|
|
|
self.verbosity = verbosity |
|
29
|
|
|
self.distribution = distribution |
|
30
|
|
|
self.n_processes = n_processes |
|
31
|
|
|
|
|
32
|
|
|
self.opt_pros = {} |
|
33
|
|
|
|
|
34
|
|
|
def _create_shared_memory(self, new_opt): |
|
35
|
|
|
if new_opt.memory == "share": |
|
36
|
|
|
if len(self.opt_pros) == 0: |
|
37
|
|
|
|
|
38
|
|
|
manager = mp.Manager() |
|
39
|
|
|
new_opt.memory = manager.dict() |
|
40
|
|
|
|
|
41
|
|
|
for opt in self.opt_pros.values(): |
|
42
|
|
|
same_obj_func = ( |
|
43
|
|
|
opt.objective_function.__name__ |
|
44
|
|
|
== new_opt.objective_function.__name__ |
|
45
|
|
|
) |
|
46
|
|
|
same_ss_length = len(opt.s_space()) == len(new_opt.s_space()) |
|
47
|
|
|
|
|
48
|
|
|
if same_obj_func and same_ss_length: |
|
49
|
|
|
new_opt.memory = opt.memory # get same manager.dict |
|
50
|
|
|
else: |
|
51
|
|
|
manager = mp.Manager() # get new manager.dict |
|
52
|
|
|
new_opt.memory = manager.dict() |
|
53
|
|
|
|
|
54
|
|
|
@staticmethod |
|
55
|
|
|
def _default_opt(optimizer): |
|
56
|
|
|
if isinstance(optimizer, str): |
|
57
|
|
|
if optimizer == "default": |
|
58
|
|
|
optimizer = RandomSearchOptimizer() |
|
59
|
|
|
return optimizer |
|
60
|
|
|
|
|
61
|
|
|
@staticmethod |
|
62
|
|
|
def _default_search_id(search_id, objective_function): |
|
63
|
|
|
if not search_id: |
|
64
|
|
|
search_id = objective_function.__name__ |
|
65
|
|
|
return search_id |
|
66
|
|
|
|
|
67
|
|
|
@staticmethod |
|
68
|
|
|
def check_list(search_space): |
|
69
|
|
|
for key in search_space.keys(): |
|
70
|
|
|
search_dim = search_space[key] |
|
71
|
|
|
|
|
72
|
|
|
error_msg = ( |
|
73
|
|
|
"Value in '{}' of search space dictionary must be of type list".format( |
|
74
|
|
|
key |
|
75
|
|
|
) |
|
76
|
|
|
) |
|
77
|
|
|
if not isinstance(search_dim, list): |
|
78
|
|
|
print("Warning", error_msg) |
|
79
|
|
|
# raise ValueError(error_msg) |
|
80
|
|
|
|
|
81
|
|
|
def add_search( |
|
82
|
|
|
self, |
|
83
|
|
|
objective_function, |
|
84
|
|
|
search_space, |
|
85
|
|
|
n_iter, |
|
86
|
|
|
search_id=None, |
|
87
|
|
|
optimizer="default", |
|
88
|
|
|
n_jobs=1, |
|
89
|
|
|
initialize={"grid": 4, "random": 2, "vertices": 4}, |
|
90
|
|
|
pass_through={}, |
|
91
|
|
|
max_score=None, |
|
92
|
|
|
early_stopping=None, |
|
93
|
|
|
random_state=None, |
|
94
|
|
|
memory="share", |
|
95
|
|
|
memory_warm_start=None, |
|
96
|
|
|
): |
|
97
|
|
|
self.check_list(search_space) |
|
98
|
|
|
|
|
99
|
|
|
optimizer = self._default_opt(optimizer) |
|
100
|
|
|
search_id = self._default_search_id(search_id, objective_function) |
|
101
|
|
|
s_space = SearchSpace(search_space) |
|
102
|
|
|
|
|
103
|
|
|
optimizer.setup_search( |
|
104
|
|
|
objective_function, |
|
105
|
|
|
s_space, |
|
106
|
|
|
n_iter, |
|
107
|
|
|
initialize, |
|
108
|
|
|
pass_through, |
|
109
|
|
|
max_score, |
|
110
|
|
|
early_stopping, |
|
111
|
|
|
random_state, |
|
112
|
|
|
memory, |
|
113
|
|
|
memory_warm_start, |
|
114
|
|
|
self.verbosity, |
|
115
|
|
|
) |
|
116
|
|
|
|
|
117
|
|
|
if memory == "share": |
|
118
|
|
|
self._create_shared_memory(optimizer) |
|
119
|
|
|
|
|
120
|
|
|
if n_jobs == -1: |
|
121
|
|
|
n_jobs = mp.cpu_count() |
|
122
|
|
|
|
|
123
|
|
|
for _ in range(n_jobs): |
|
124
|
|
|
nth_process = len(self.opt_pros) |
|
125
|
|
|
self.opt_pros[nth_process] = optimizer |
|
126
|
|
|
|
|
127
|
|
|
def _print_info(self): |
|
128
|
|
|
print_res = PrintResults(self.opt_pros, self.verbosity) |
|
129
|
|
|
|
|
130
|
|
|
for results in self.results_list: |
|
131
|
|
|
nth_process = results["nth_process"] |
|
132
|
|
|
print_res.print_process(results, nth_process) |
|
133
|
|
|
|
|
134
|
|
|
def run(self, max_time=None): |
|
135
|
|
|
for opt in self.opt_pros.values(): |
|
136
|
|
|
opt.max_time = max_time |
|
137
|
|
|
|
|
138
|
|
|
self.results_list = run_search( |
|
139
|
|
|
self.opt_pros, self.distribution, self.n_processes |
|
140
|
|
|
) |
|
141
|
|
|
|
|
142
|
|
|
self.results_ = Results(self.results_list, self.opt_pros) |
|
143
|
|
|
|
|
144
|
|
|
self._print_info() |
|
145
|
|
|
|
|
146
|
|
|
def best_para(self, id_): |
|
147
|
|
|
return self.results_.best_para(id_) |
|
148
|
|
|
|
|
149
|
|
|
def best_score(self, id_): |
|
150
|
|
|
return self.results_.best_score(id_) |
|
151
|
|
|
|
|
152
|
|
|
def search_data(self, id_): |
|
153
|
|
|
return self.results_.search_data(id_) |
|
154
|
|
|
|
SimonBlanke /
Hyperactive
