-
Notifications
You must be signed in to change notification settings - Fork 0
/
test_execute.py
181 lines (136 loc) · 4.65 KB
/
test_execute.py
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
168
169
170
171
172
173
174
175
176
177
178
179
180
181
# coding: utf-8
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
import unittest
import os
# import shutil
import numpy as np
from saep.test_utils_const import FIXED_SEED
from dataset import FEATURES_KEY, establish_baselines
from hparam import super_input_fn
from execute import utilise_AdaNet, utilise_SAEP
from execute import ensemble_pruning_set, auxrun_expts
# --------------------------------------
prng = np.random.RandomState(FIXED_SEED)
RANDOM_SEED = 10000 - FIXED_SEED
LEARNING_RATE = 0.001
TRAIN_STEPS = 2 # 6000
BATCH_SIZE = 2
ADANET_ITERATIONS = 2 # 11
ADANET_LAMBDA = 0
LEARN_MIXTURE_WEIGHTS = False
LOG_TLE = 'discard'
LOG_DIR = os.path.join(os.getcwd(), "tmpmodels", "discard")
TF_LOG_TLE = 'discard_tf'
logger = None
nb_feat, nb_labl = 7, 4
# nb_trn, nb_tst = 20, 4
nb_trn, nb_tst = 10, 2
nb_shap = (28, 28, 1)
y_trn = prng.randint(nb_labl, size=nb_trn)
y_tst = prng.randint(nb_labl, size=nb_tst)
X_trn = prng.rand(nb_trn, *nb_shap)
X_tst = prng.rand(nb_tst, *nb_shap)
this_experiment = 'casual'
def _run(tp, lmw, um):
experiment_name, this_experiment = auxrun_expts(
tp, lmw=lmw, modeluse=um)
directory = os.path.join(LOG_DIR, this_experiment)
# if os.path.exists(directory):
# shutil.rmtree(directory)
input_fn = super_input_fn(
X_trn, y_trn, X_tst, y_tst, nb_shap, RANDOM_SEED)
head, feature_columns = establish_baselines(
nb_labl, nb_shap, FEATURES_KEY)
return experiment_name, this_experiment, \
input_fn, head, feature_columns
# --------------------------------------
class Test_SAEP(unittest.TestCase):
def curr(self, tp, lmw=False, um='dnn'):
if tp.endswith('O'):
cs = utilise_SAEP(tp, modeluse=um)
else:
cs = utilise_SAEP(
tp, learn_mixture_weights=lmw, modeluse=um)
exnm, excr, input_fn, head, fcs = _run(tp, lmw, um)
cs.assign_expt_params(nb_labl, excr, LOG_DIR)
cs.assign_train_param(LEARNING_RATE, BATCH_SIZE, TRAIN_STEPS)
cs.assign_adanet_para(ADANET_ITERATIONS, ADANET_LAMBDA)
ep = tp[:-2]
cs.assign_SAEP_adapru(ensemble_pruning_set[ep])
cs.assign_SAEP_logger(logger)
et = cs.create_estimator(um, fcs, head, input_fn)
# r, et = cs.train_and_evaluate(et, input_fn)
def impl(self, tp, lmw=False):
if tp.endswith('O'):
self.curr(tp, um='dnn')
self.curr(tp, um='cnn')
# self.curr(tp, um='cpx')
return
self.curr(tp, lmw, 'dnn')
self.curr(tp, lmw, 'cnn')
# self.curr(tp, lmw, 'cpx')
def test_main(self):
self.impl('SAEP.O')
# self.impl('SAEP.W', lmw=False)
# self.impl('SAEP.W', lmw=True)
self.impl('AdaNet.O')
self.impl('AdaNet.W', lmw=False)
self.impl('AdaNet.W', lmw=True)
class Test_PRS(Test_SAEP):
def test_main(self):
self.impl('PRS.O')
self.impl('PRS.W', lmw=False)
self.impl('PRS.W', lmw=True)
class Test_PAP(Test_SAEP):
def test_main(self):
self.impl('PAP.O')
self.impl('PAP.W', lmw=False)
self.impl('PAP.W', lmw=True)
class Test_PIE(Test_SAEP):
def curr(self, tp, lmw=False, um='dnn'):
if tp.endswith('O'):
cs = utilise_SAEP(tp, 0.4, modeluse=um)
else:
cs = utilise_SAEP(tp, 0.4, lmw, um)
_, exnm, input_fn, head, fcs = _run(tp, lmw, um)
cs.assign_expt_params(nb_labl, exnm, LOG_DIR)
cs.assign_train_param(LEARNING_RATE, BATCH_SIZE, TRAIN_STEPS)
cs.assign_adanet_para(ADANET_ITERATIONS, ADANET_LAMBDA)
ep = tp[:-2]
cs.assign_SAEP_adapru(ensemble_pruning_set[ep], thinp_alpha=.4)
cs.assign_SAEP_logger(logger)
et = cs.create_estimator(um, fcs, head, input_fn)
def test_main(self):
self.impl('PIE.O')
self.impl('PIE.W', lmw=False)
self.impl('PIE.W', lmw=True)
# --------------------------------------
class Test_adanet(unittest.TestCase):
def curr(self, tp, lmw=False, um='linear'):
if tp.endswith('O'):
cs = utilise_AdaNet(tp, modeluse=um)
else:
cs = utilise_AdaNet(tp, lmw, um)
ex_nm, ex_cr, input_fn, head, fcs = _run(tp, lmw, um)
cs.assign_expt_params(nb_labl, ex_cr, LOG_DIR)
cs.assign_train_param(LEARNING_RATE, BATCH_SIZE, TRAIN_STEPS)
cs.assign_adanet_para(ADANET_ITERATIONS, ADANET_LAMBDA)
et = cs.create_estimator(um, fcs, head, input_fn)
# if um == 'linear':
# return
# input_fn = _run(tp, lmw, um)
# r, et = cs.train_and_evaluate(et, input_fn)
def impl(self, tp, lmw=False):
if tp.endswith('O'):
self.curr(tp, um='dnn')
self.curr(tp, um='cnn')
self.curr(tp, um='linear')
return
self.curr(tp, lmw, 'dnn')
self.curr(tp, lmw, 'cnn')
def test_main(self):
self.impl('AdaNet.O')
self.impl('AdaNet.W', False)
self.impl('AdaNet.W', True)