FP-GNN/hyper_opti.py at main · idrugLab/FP-GNN · 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
from argparse import ArgumentParser, Namespace
from hyperopt import fmin, tpe, hp
import numpy as np
import os
from copy import deepcopy
from fpgnn.tool import set_hyper_argument, set_log
from train import training

space = {
         'fp_2_dim':hp.quniform('fp_2_dim', low=300, high=600, q=50),
         'nhid':hp.quniform('nhid', low=40, high=80, q=5),
         'nheads':hp.quniform('nheads', low=2, high=8, q=1),
         'gat_scale':hp.quniform('gat_scale', low=0.2, high=0.8, q=0.1),
         'dropout':hp.quniform('dropout', low=0.0, high=0.6, q=0.05),
         'dropout_gat':hp.quniform('dropout_gat', low=0.0, high=0.6, q=0.05)
}


def fn(space):
    search_no = args.search_now
    log_name = 'train'+str(search_no)
    log = set_log(log_name,args.log_path)
    result_path = os.path.join(args.log_path, 'hyper_para_result.txt')

    list = ['fp_2_dim','nhid','nheads']
    for one in list:
        space[one] = int(space[one])
    hyperp = deepcopy(args)
    name_list = []
    change_args = []
    for key,value in space.items():
        name_list.append(str(key))
        name_list.append('-')
        name_list.append((str(value))[:5])
        name_list.append('-')
        setattr(hyperp,key,value)
    dir_name = "".join(name_list)
    dir_name = dir_name[:-1]
    hyperp.save_path = os.path.join(hyperp.save_path, dir_name)

    ave,std = training(hyperp,log)

    with open(result_path,'a') as file:
        file.write(str(space)+'\n')
        file.write('Result '+str(hyperp.metric)+' : '+str(ave)+' +/- '+str(std)+'\n')

    if ave is None:
        if hyperp.dataset_type == 'classification':
            ave = 0
        else:
            raise ValueError('Result of model is error.')

    args.search_now += 1

    if hyperp.dataset_type == 'classification':
        return -ave
    else:
        return ave

def hyper_searching(args):
    result_path = os.path.join(args.log_path, 'hyper_para_result.txt')

    result = fmin(fn,space,tpe.suggest,args.search_num)

    with open(result_path,'a') as file:
        file.write('Best Hyperparameters : \n')
        file.write(str(result)+'\n')


if __name__ == '__main__':
    args = set_hyper_argument()
    hyper_searching(args)