-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathutils.py
234 lines (210 loc) · 11.6 KB
/
utils.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
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
import math
import torch
import numpy as np
from config import Config
from tensorboard_logger import Logger as TbLogger
import os
def log_values(val_accuracy, val_loss, val_kacc, epoch, log_path, problem, LEARNING_RATE):
with open(os.path.join(log_path, f'{problem}_trace.txt'), 'a+') as f:
f.writelines(f'epoch: {epoch}, val_accuracy: {val_accuracy}, '
f'val_loss: {val_loss}, kacc: {val_kacc[0]}, {val_kacc[1]}, {val_kacc[2]}'
+ f' lr{LEARNING_RATE}'+ '\r\n')
# Log values to tensorboard
tb_logger = TbLogger(log_path)
tb_logger.log_value('val_accuracy', val_accuracy, epoch)
tb_logger.log_value('val_loss', val_loss, epoch)
def tensor(x):
if isinstance(x, torch.Tensor):
return x
x = np.asarray(x, dtype=np.float32)
x = torch.from_numpy(x).to(Config.DEVICE)
return x
def to_np(t):
return t.cpu().detach().numpy()
class SCIPParam:
def __init__(self, init_param=None):
if init_param is None:
init_param = {}
self.params = init_param
def setIntParam(self, key, value):
self.params[key] = value
def getParam(self):
return self.params
def init_params(init_param = None, presolve = False, change_default_settings = False, disable_all_h = False):
"""
:param model: scip.Model(), model instantiation
:param scip_limits: dict, specifying SCIP parameter limits
:param scip_params: dict, specifying SCIP parameter setting
:return: -
Initialize SCIP parameters for the model.
"""
model = SCIPParam(init_param)
model.setIntParam('display/verblevel', 0)
# limits
model.setIntParam('limits/nodes', -1)
model.setIntParam('limits/time', 3600.0)
if not presolve:
# disable presolve and cuts (enabled in default)
model.setIntParam('presolving/maxrounds', 0) # 0: off, -1: unlimited
model.setIntParam('separating/maxrounds', 0) # 0 to disable local separation
model.setIntParam('separating/maxroundsroot', 0) # 0 to disable root separation
model.setIntParam('presolving/maxrestarts', 0)
if change_default_settings:
scip_params = {'heuristics': False, # enable all primal heuristics
'cutoff': True, # provide cutoff (value needs to be passed to the environment)
'conflict_usesb': False, # use SB conflict analysis
'probing_bounds': False, # use probing bounds identified during SB
'checksol': False,
'reevalage': 0,
}
else:
scip_params ={
'heuristics': True,
'cutoff': False,
'conflict_usesb': True,
'probing_bounds': True,
'checksol': True,
'reevalage': 10,
}
# # disable reoptimization (as in default)
# model.setIntParam('reoptimization/enable', False)
#
# # cutoff value is eventually set in env.run_episode
# # other parameters to be disabled in 'sandbox' setting
# model.setIntParam('conflict/usesb', scip_params['conflict_usesb'])
# model.setIntParam('branching/fullstrong/probingbounds', scip_params['probing_bounds'])
# model.setIntParam('branching/relpscost/probingbounds', scip_params['probing_bounds'])
# model.setIntParam('branching/checksol', scip_params['checksol'])
# model.setIntParam('branching/fullstrong/reevalage', scip_params['reevalage'])
# primal heuristics (54 total, 14 of which are disabled in default setting as well)
if not scip_params['heuristics']:
model.setIntParam('heuristics/actconsdiving/freq', -1) # disabled at default
model.setIntParam('heuristics/bound/freq', -1) # disabled at default
model.setIntParam('heuristics/clique/freq', -1)
model.setIntParam('heuristics/coefdiving/freq', -1)
model.setIntParam('heuristics/completesol/freq', -1)
model.setIntParam('heuristics/conflictdiving/freq', -1) # disabled at default
model.setIntParam('heuristics/crossover/freq', -1)
model.setIntParam('heuristics/dins/freq', -1) # disabled at default
model.setIntParam('heuristics/distributiondiving/freq', -1)
# model.setIntParam('heuristics/dualval/freq', -1) # disabled at default
model.setIntParam('heuristics/farkasdiving/freq', -1)
model.setIntParam('heuristics/feaspump/freq', -1)
model.setIntParam('heuristics/fixandinfer/freq', -1) # disabled at default
model.setIntParam('heuristics/fracdiving/freq', -1)
model.setIntParam('heuristics/gins/freq', -1)
model.setIntParam('heuristics/guideddiving/freq', -1)
model.setIntParam('heuristics/zeroobj/freq', -1) # disabled at default
model.setIntParam('heuristics/indicator/freq', -1)
model.setIntParam('heuristics/intdiving/freq', -1) # disabled at default
model.setIntParam('heuristics/intshifting/freq', -1)
model.setIntParam('heuristics/linesearchdiving/freq', -1)
model.setIntParam('heuristics/localbranching/freq', -1) # disabled at default
model.setIntParam('heuristics/locks/freq', -1)
model.setIntParam('heuristics/lpface/freq', -1)
model.setIntParam('heuristics/alns/freq', -1)
model.setIntParam('heuristics/nlpdiving/freq', -1)
model.setIntParam('heuristics/mutation/freq', -1) # disabled at default
model.setIntParam('heuristics/multistart/freq', -1)
model.setIntParam('heuristics/mpec/freq', -1)
model.setIntParam('heuristics/objpscostdiving/freq', -1)
model.setIntParam('heuristics/octane/freq', -1) # disabled at default
model.setIntParam('heuristics/ofins/freq', -1)
model.setIntParam('heuristics/oneopt/freq', -1)
model.setIntParam('heuristics/proximity/freq', -1) # disabled at default
model.setIntParam('heuristics/pscostdiving/freq', -1)
model.setIntParam('heuristics/randrounding/freq', -1)
model.setIntParam('heuristics/rens/freq', -1)
model.setIntParam('heuristics/reoptsols/freq', -1)
model.setIntParam('heuristics/repair/freq', -1) # disabled at default
model.setIntParam('heuristics/rins/freq', -1)
model.setIntParam('heuristics/rootsoldiving/freq', -1)
model.setIntParam('heuristics/rounding/freq', -1)
model.setIntParam('heuristics/shiftandpropagate/freq', -1)
model.setIntParam('heuristics/shifting/freq', -1)
model.setIntParam('heuristics/simplerounding/freq', -1)
model.setIntParam('heuristics/subnlp/freq', -1)
model.setIntParam('heuristics/trivial/freq', -1)
model.setIntParam('heuristics/trivialnegation/freq', -1)
model.setIntParam('heuristics/trysol/freq', -1)
model.setIntParam('heuristics/twoopt/freq', -1) # disabled at default
model.setIntParam('heuristics/undercover/freq', -1)
model.setIntParam('heuristics/vbounds/freq', -1)
model.setIntParam('heuristics/veclendiving/freq', -1)
model.setIntParam('heuristics/zirounding/freq', -1)
if disable_all_h:
model.setIntParam('heuristics/actconsdiving/freq', 0) # disabled at default
model.setIntParam('heuristics/bound/freq', 0) # disabled at default
model.setIntParam('heuristics/clique/freq', 0)
model.setIntParam('heuristics/coefdiving/freq', 0)
model.setIntParam('heuristics/completesol/freq', 0)
model.setIntParam('heuristics/conflictdiving/freq', 0) # disabled at default
model.setIntParam('heuristics/crossover/freq', 0)
model.setIntParam('heuristics/dins/freq', 0) # disabled at default
model.setIntParam('heuristics/distributiondiving/freq', 0)
# model.setIntParam('heuristics/dualval/freq', 0) # disabled at default
model.setIntParam('heuristics/farkasdiving/freq', 0)
model.setIntParam('heuristics/feaspump/freq', 0)
model.setIntParam('heuristics/fixandinfer/freq', 0) # disabled at default
model.setIntParam('heuristics/fracdiving/freq', 0)
model.setIntParam('heuristics/gins/freq', 0)
model.setIntParam('heuristics/guideddiving/freq', 0)
model.setIntParam('heuristics/zeroobj/freq', 0) # disabled at default
model.setIntParam('heuristics/indicator/freq', 0)
model.setIntParam('heuristics/intdiving/freq', 0) # disabled at default
model.setIntParam('heuristics/intshifting/freq', 0)
model.setIntParam('heuristics/linesearchdiving/freq', 0)
model.setIntParam('heuristics/localbranching/freq', 0) # disabled at default
model.setIntParam('heuristics/locks/freq', 0)
model.setIntParam('heuristics/lpface/freq', 0)
model.setIntParam('heuristics/alns/freq', 0)
model.setIntParam('heuristics/nlpdiving/freq', 0)
model.setIntParam('heuristics/mutation/freq', 0) # disabled at default
model.setIntParam('heuristics/multistart/freq', 0)
model.setIntParam('heuristics/mpec/freq', 0)
model.setIntParam('heuristics/objpscostdiving/freq', 0)
model.setIntParam('heuristics/octane/freq', 0) # disabled at default
model.setIntParam('heuristics/ofins/freq', 0)
model.setIntParam('heuristics/oneopt/freq', 0)
model.setIntParam('heuristics/proximity/freq', 0) # disabled at default
model.setIntParam('heuristics/pscostdiving/freq', 0)
model.setIntParam('heuristics/randrounding/freq', 0)
model.setIntParam('heuristics/rens/freq', 0)
model.setIntParam('heuristics/reoptsols/freq', 0)
# model.setIntParam('heuristics/repair/freq', 0) # disabled at default
model.setIntParam('heuristics/rins/freq', 0)
model.setIntParam('heuristics/rootsoldiving/freq', 0)
model.setIntParam('heuristics/rounding/freq', 0)
model.setIntParam('heuristics/shiftandpropagate/freq', 0)
model.setIntParam('heuristics/shifting/freq', 0)
model.setIntParam('heuristics/simplerounding/freq', 0)
model.setIntParam('heuristics/subnlp/freq', 0)
model.setIntParam('heuristics/trivial/freq', 0)
model.setIntParam('heuristics/trivialnegation/freq', 0)
model.setIntParam('heuristics/trysol/freq', 0)
model.setIntParam('heuristics/twoopt/freq', 0) # disabled at default
model.setIntParam('heuristics/undercover/freq', 0)
model.setIntParam('heuristics/vbounds/freq', 0)
model.setIntParam('heuristics/veclendiving/freq', 0)
model.setIntParam('heuristics/zirounding/freq', 0)
return model.getParam()
def clip_grad_norms(param_groups, max_norm=math.inf):
"""
Clips the norms for all param groups to max_norm and returns gradient norms before clipping
:param optimizer:
:param max_norm:
:param gradient_norms_log:
:return: grad_norms, clipped_grad_norms: list with (clipped) gradient norms per group
"""
grad_norms = [
torch.nn.utils.clip_grad_norm_(
group['params'],
max_norm if max_norm > 0 else math.inf, # Inf so no clipping but still call to calc
norm_type=2
)
for group in param_groups
]
grad_norms_clipped = [min(g_norm, max_norm) for g_norm in grad_norms] if max_norm > 0 else grad_norms
return grad_norms, grad_norms_clipped
if __name__ == '__main__':
print(init_params().__len__())