-
Notifications
You must be signed in to change notification settings - Fork 0
/
parameters.py
211 lines (178 loc) · 4.63 KB
/
parameters.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
"""
Hyper-parameters
"""
import argparse
def get_arguments():
""" Hyper-parameters """
parser = argparse.ArgumentParser(
description="Train Polarization_quantization_model"
)
parser.add_argument(
"--data_dir",
type=str,
default="./datasets",
metavar="datapath",
help="Location of data files (datasets)",
)
parser.add_argument(
"--epochs",
type=int,
default=60,
metavar="N",
help="number of epochs to train (default: 60)",
)
parser.add_argument(
"--lr",
type=float,
default=1e-3,
metavar="LR",
help="learning rate (default: 0.001)",
)
parser.add_argument("--decay", type=float, default=0, help="weight decay")
parser.add_argument(
"--save_dir",
type=str,
default="./checkpoints/",
metavar="savepath",
help="save path",
)
parser.add_argument(
"--checkpoint_name",
type=str,
default="polarized_model",
metavar="name",
help="name of checkpoint",
)
parser.add_argument(
"--dataset",
type=str,
default="fashion",
metavar="mnist/fashion",
help="Which dataset to use (default: fashion)",
)
parser.add_argument(
"--model",
type=str,
default="Polarization_quantization_model",
help="Which model model to train (default: Polarization_quantization_model)",
)
parser.add_argument(
"--bpda_steepness",
type=int,
default=8,
metavar="bs",
help="backward pass differentiable approximation steepness (default: 8)",
)
parser.add_argument(
"--jump",
type=float,
default=0.2,
metavar="jump",
help="jump of saturation activation function (default: 0.2)",
)
parser.add_argument(
"--batch_size",
type=int,
default=64,
metavar="LR",
help="batch size (default: 64)",
)
parser.add_argument(
"--num_ckpt_steps",
type=int,
default=10,
help="save checkpoint steps (default: 10)",
)
parser.add_argument(
"--resume", "-r", action="store_true", help="resume from checkpoint"
)
parser.add_argument(
"--savefig_attack",
action="store_true",
help="plot correctly and wrongly classified images after attack",
)
parser.add_argument(
"--savefig_train",
action="store_true",
help="plot filters and histograms in each epoch of training",
)
parser.add_argument(
"--bs1", type=float, default=1, metavar="bump_scale1", help="bump scale stage 1"
)
parser.add_argument(
"--bw1",
type=float,
default=0.35,
metavar="bump_width1",
help="bump width stage 1",
)
parser.add_argument(
"--bs2", type=float, default=1, metavar="bump_scale2", help="bump scale stage 2"
)
parser.add_argument(
"--bw2",
type=float,
default=0.15,
metavar="bump_width2",
help="bump width stage 2",
)
parser.add_argument(
"--stage2_start",
type=int,
default=20,
metavar="stage2_epoch",
help="Start epoch of stage 2",
)
parser.add_argument(
"--stage3_start",
type=int,
default=40,
metavar="stage3_epoch",
help="Start epoch of stage 3",
)
parser.add_argument(
"--attack_method",
type=str,
default="pgd",
metavar="fgsm/pgd/r_iter",
help="Attack method to be used",
)
parser.add_argument(
"--attack_batch_size",
type=int,
default=1000,
metavar="LR",
help="batch size (default: 64)",
)
parser.add_argument(
"--save_attack",
action="store_true",
default=False,
help="whether to save attack images",
)
parser.add_argument(
"--eps", type=float, default=0.1, metavar="eps", help="Attack budget epsilon"
)
parser.add_argument(
"--step_size",
type=float,
default=0.01,
metavar="eps-itr",
help="(Iterative attacks) Attack budget in each iteration",
)
parser.add_argument(
"--num_steps",
type=int,
default=100,
metavar="N",
help="(Iterative attacks) Number of iterations in attack",
)
parser.add_argument(
"--num_restarts",
type=int,
default=20,
metavar="N",
help="(PGD) Number of random restarts in attack",
)
args = parser.parse_args()
return args