DRepL-Project-2023/bay_hyperparam_search.py at main · madammann/DRepL-Project-2023 · GitHub

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
import os
import argparse

import numpy as np
import pandas as pd

from training import train

parser = argparse.ArgumentParser(description='empty')
parser.add_argument('--iter',default='100')
parser.add_argument('--path',default='./hyperparams/hyperparams.csv')
parser.add_argument('--cnn',default='False')

DEFAULT_PARAMS = {
    'batches_per_epoch' : 1000,
    'learning_rate' : 0.001,
    'gamma' : 0.9,
    'epsilon' : 0.9,
    'epsilon_decay' : 0.1,
    'buffer_size_in_batches' : 1000,
    'batch_size' : 64,
    'replay_ratio' : 0.1
}

STATIC_MLP = {
    'epochs' : 1000,
    'visual' : True,
    'rgb' : True,
    'cnn_depth' : 1,
    'mlp_layers' : 1,
    'head_layers' : 1,
    'filters' : 1,
    'kernel_size' : 3,
    'k_init' : 'glorot_uniform',
    'b_init' : 'zeros'
}

STATIC_CNN = {
    'epochs' : 1000,
    'visual' : True,
    'rgb' : True,
    'cnn_depth' : 1,
    'mlp_layers' : 1,
    'head_layers' : 1,
    'filters' : 1,
    'kernel_size' : 3,
    'k_init' : 'glorot_uniform',
    'b_init' : 'zeros'
}

SEARCH_SPACE_BORDERS = {
    'learning_rate' : (0.001, 0.2),
    'gamma' : (0.1, 0.99),
    'epsilon' : (0.0, 0.99),
    'epsilon_decay' : (0.0, 0.99),
    'buffer_size_in_batches' : (1000, 10000),
    'batch_size' : [1, 16, 32, 64],
    'replay_ratio' : (0.01, 0.99)
}

#for search keep initial random buffer(s) in memory/drive between searches

class CustomBayesianHyperparameterOptimizer:
    '''
    ADD
    '''

    def __init__(self, training_func, export_path : str, params : dict, restraints : dict, static_params : dict):
        '''
        ADD
        '''

        self.export_path = export_path
        self.training_func = training_func
        self.params = params
        self.restraints = restraints

        self.data = [] if not os.path.exists(self.export_path) else self.load_dataset()

    def find_most_promising_param_set(self):
        '''
        ADD
        '''

        pass

    def search(self, epoch_depth=1):
        '''
        ADD
        '''

        pass

    def store_dataset(self):
        '''
        ADD
        '''

        pass

    def load_dataset(self):
        '''
        ADD
        '''

        pass

    def __call__(self, iter_lim=100):
        '''
        ADD
        '''

        pass


# we try to find the paramters theta which maximize the objective function

def main(train, path):
    CustomBayesianHyperparameterOptimizer(train, path, params, restraints, static_params)

if __name__ == '__main__':
    pass