Recommendation_Algos/run_test_opt.py at master · leonvking0/Recommendation_Algos · 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
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
import time
import random
import numpy as np
from algorithms.gru4rec import gru4rec2
from evaluation import evaluation as eval
from evaluation import loader as loader
from evaluation.metrics import accuracy as ac
from evaluation.metrics import coverage as cov
from evaluation.metrics import popularity as pop
import theano
import theano.misc.pkl_utils as pkl
import sys
import traceback
import gc


if __name__ == '__main__':

    '''
    Configuration
    '''
    data_path = 'data/rsc15/single/'
    file_prefix = 'rsc15-clicks'

    limit_train = None  # limit in number of rows or None
    limit_test = None  # limit in number of rows or None
    density_value = 1  # randomly filter out events (0.0-1.0, 1:keep all)

    export_csv_base = 'results/run-test-rsc15-opt'

    learn_rates = np.linspace(0.05, 0.8, num=16).astype( theano.config.floatX )
    drop_outs = np.linspace(0.0, 0.9, num=10).astype( theano.config.floatX )
    momentums = np.linspace(0.0, 0.9, num=10).astype( theano.config.floatX )
    losses = ['bpr-max-0.5', 'top1-max']

    best = 0.0
    best_key = ''

    for test_num in range(33):

        try:

            print('process test ' + str(test_num))

            lr = random.choice(learn_rates)
            drop = random.choice(drop_outs)
            momentum = random.choice(momentums)
            loss = random.choice(losses)

            export_csv = export_csv_base + str(test_num) + '.csv'

            # create a list of metric classes to be evaluated
            metric = []
            metric.append(ac.HitRate(20))
            metric.append(ac.HitRate(10))
            metric.append(ac.HitRate(3))
            metric.append(ac.MRR(20))
            metric.append(ac.MRR(10))
            metric.append(ac.MRR(3))
            metric.append(cov.Coverage(20))
            metric.append(pop.Popularity(20))

            # create a dict of (textual algorithm description => class) to be evaluated
            algs = {}

            key = 'gru4rec2-' + loss + '-lr' + str(lr) + '-do' + str(drop) + '-mom' + str(momentum) + 't' + str(test_num)
            print('TESTING: ' + key)
            drop=float(drop)

            gru = gru4rec2.GRU4Rec(loss=loss, final_act='linear', hidden_act='tanh', layers=[100], batch_size=64, dropout_p_hidden=drop, learning_rate=lr, momentum=momentum, n_epochs=10,  n_sample=2048, sample_alpha=0, time_sort=True)
            algs[key] = gru

            '''
            Execution
            '''
            # load data

            print('data_path: ', data_path)

            train, test = loader.load_data(data_path, file_prefix, rows_train=limit_train, rows_test=limit_test, density=density_value, slice_num=slice)

            item_ids = train.ItemId.unique()

            # init metrics
            for m in metric:
                m.init(train)

            # train algorithms
            for k, a in algs.items():
                ts = time.time()
                print('fit ', k)
                a.fit(train)
                print(k, ' time: ', (time.time() - ts))

            # result dict
            res = {};

            # evaluation
            for k, a in algs.items():
                try:
                    res[k] = eval.evaluate_sessions(a, metric, test, train)
                    a.clear()
                except Exception:
                    a.clear()
                    del a
                    del algs
                    del metric
                    del train
                    del test
                    del gru
                    print('cleared gru with nan error')
                    raise

            # print results
            for k, l in res.items():
                for e in l:
                    print(k, ':', e[0], ' ', e[1])

            if export_csv is not None:

                fileH = open(export_csv, 'w+')
                fileH.write('Metrics;')

                for k, l in res.items():
                    for e in l:
                        fileH.write(e[0])
                        fileH.write(';')
                    break

                fileH.write('\n')

                for k, l in res.items():
                    fileH.write(k)
                    fileH.write(';')
                    for e in l:
                        fileH.write(str(e[1]))
                        fileH.write(';')
                    fileH.write('\n')
                fileH.close()

            if res[key][0][1] > best:  # new best found
                best = res[key][0][1]
                best_key = key

            print('CURRENT BEST: ' + best_key)
            print('WITH HR@20: ' + str(best))

            del drop
            del momentum
            del loss
            del lr
            del algs
            del metric
            del train
            del test
            del gru
            del res
            del item_ids
            del fileH

            for i in range(5): gc.collect()


        except Exception:

            traceback.print_exc()