-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathconfig.py
169 lines (162 loc) · 10.3 KB
/
config.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
import argparse
import models
from data import valid_datasets as dataset_names
model_names = sorted(name for name in models.__dict__
if name.islower() and not name.startswith("__")
and callable(models.__dict__[name]))
r'''version list
- v1
- v2
- v2
- v2q ($\alpha$, $\beta$ quantization and adding epsilon in denom of $\alpha$)
- v2qq (quantization ver with $\alpha$, $\beta$ quantization and adding epsilon in denom of $\alpha$)
- v2f (fixed index $k$ during retraining time with v2qq)
- v2qnb (no $\beta$ with v2q)
- v2qqnb (no $\beta$ with v2qq)
- v2.5 (old: v3)
- v3: rotation, flip, shift
'''
versions = [
'v1',
'v2', 'v2q', 'v2qq', 'v2f', 'v2nb', 'v2qnb', 'v2qqnb',
# 'v2.5',
# 'v3',
]
# sacred setting
MONGO_URI = 'mongodb://mlvc:mlvcdatabase!@mlvc.khu.ac.kr:31912'
MONGO_DB = 'training'
def config():
r"""configuration settings
"""
parser = argparse.ArgumentParser(description='KH Research')
parser.add_argument('dataset', metavar='DATA', default='cifar10',
choices=dataset_names,
help='dataset: ' +
' | '.join(dataset_names) +
' (default: cifar10)')
parser.add_argument('-a', '--arch', metavar='ARCH', default='mobilenet',
choices=model_names,
help='model architecture: ' +
' | '.join(model_names) +
' (default: mobilenet)')
parser.add_argument('--layers', default=16, type=int, metavar='N',
help='number of layers in VGG/ResNet/WideResNet (default: 16)')
parser.add_argument('--bn', '--batch-norm', dest='bn', action='store_true',
help='Use batch norm in VGG?')
parser.add_argument('--width-mult', default=1.0, type=float, metavar='WM',
help='width multiplier to thin a network '
'uniformly at each layer (default: 1.0)')
parser.add_argument('--dr', '--drop-rate', default=0.3, type=float,
metavar='DR', help='dropout rate for WRN (default: 0.3)',
dest='drop_rate')
parser.add_argument('-j', '--workers', default=8, type=int, metavar='N',
help='number of data loading workers (default: 8)')
parser.add_argument('--epochs', default=200, type=int, metavar='N',
help='number of total epochs to run (default: 200)')
parser.add_argument('-b', '--batch-size', default=256, type=int, metavar='N',
help='mini-batch size (default: 256), this is the total '
'batch size of all GPUs on the current node when '
'using Data Parallel')
parser.add_argument('--lr', '--learning-rate', default=0.1, type=float,
metavar='LR', help='initial learning rate (default: 0.1)',
dest='lr')
parser.add_argument('--momentum', default=0.9, type=float, metavar='M',
help='momentum (default: 0.9)')
parser.add_argument('--wd', '--weight-decay', default=5e-4, type=float,
metavar='W', help='weight decay (default: 5e-4)',
dest='weight_decay')
# parser.add_argument('--nesterov', dest='nesterov', action='store_true',
# help='use nesterov momentum?')
parser.add_argument('--basetest', dest='basetest', action='store_true',
help='baseline test (various weight decay test) (\'_wd\{\}\' added in filename)')
parser.add_argument('-p', '--print-freq', default=100, type=int,
metavar='N', help='print frequency (default: 100)')
parser.add_argument('--ckpt', default='', type=str, metavar='PATH',
help='Path of checkpoint for resuming/testing '
'or retraining model (Default: none)')
parser.add_argument('-R', '--resume', dest='resume', action='store_true',
help='Resume model?')
parser.add_argument('-E', '--evaluate', dest='evaluate', action='store_true',
help='Test model?')
parser.add_argument('-C', '--cuda', dest='cuda', action='store_true',
help='Use cuda?')
parser.add_argument('-T', '--retrain', dest='retrain', action='store_true',
help='Retraining?')
parser.add_argument('-g', '--gpuids', default=[0], nargs='+',
help='GPU IDs for using (Default: 0)')
parser.add_argument('--datapath', default='../data', type=str, metavar='PATH',
help='where you want to load/save your dataset? (default: ../data)')
# for new methods
parser.add_argument('-v', '--version', default='', metavar='V', dest='version',
choices=versions,
help='version: ' +
' | '.join(versions) +
' (find kernel version (default: none))')
parser.add_argument('-d', '--bind-size', default=2, type=int, metavar='N',
dest='bind_size',
help='the number of binding channels in convolution '
'(subvector size) on version 2.5 (v2.5) (default: 2)')
parser.add_argument('-pwd', '--pw-bind-size', default=8, type=int, metavar='N',
dest='pw_bind_size',
help='the number of binding channels in pointwise convolution '
'(subvector size) (default: 8)')
parser.add_argument('-pws', '--pwkernel-stride', default=1, type=int, metavar='N',
dest='pwkernel_stride',
help='the number of pwkernel stride size in reference layer '
'(default: 1)')
parser.add_argument('-N', '--new', dest='new', action='store_true',
help='new method?')
parser.add_argument('-refnum', '--ref-layer-num', dest='refnum', default=0, type=int, metavar='N',
help='reference layer number (default: 0)')
parser.add_argument('-eps', '--epsilon', dest='epsilon', default=1e-08, type=float, metavar='EPS',
help='epsilon for denominator of alpha in find_kernel (default: 1e-08)')
parser.add_argument('-s', '--save-epoch', dest='save_epoch', default=1, type=int, metavar='N',
help='number of epochs to save checkpoint and to apply new method (default: 1)')
parser.add_argument('-ustv1', '--using-sig-tanhv1', dest='ustv1', default='', type=str,
help='using sigmoid or tanh for finding similar kernel (default: none) (available options: sigmoid, tanh)')
parser.add_argument('-ustv2', '--using-sig-tanhv2', dest='ustv2', default='', type=str,
help='using sigmoid or tanh for restoration kernel (default: none) (available options: sigmoid, tanh)')
parser.add_argument('--tvl', '--tv-loss', dest='tv_loss', action='store_true',
help='total variation loss?')
parser.add_argument('--tvls', '--tvl-scale', dest='tvls', default=1e-08, type=float,
help='scale factor of tv_loss (default: 1e-08)')
parser.add_argument('--orthol', '--ortho-loss', dest='ortho_loss', action='store_true',
help='orthogonal loss?')
parser.add_argument('--orthols', '--orthol-scale', dest='orthols', default=1e-05, type=float,
help='scale factor of ortho_loss (default: 1e-05)')
parser.add_argument('--corl', '--cor-loss', dest='cor_loss', action='store_true',
help='correlation loss?')
parser.add_argument('--corls', '--corl-scale', dest='corls', default=1e-05, type=float,
help='scale factor of cor_loss (default: 1e-05)')
parser.add_argument('--orthocorl', '--ortho-cor-loss', dest='ortho_cor_loss', action='store_true',
help='orthogonal correlation loss?')
parser.add_argument('--orthocorls', '--orthocorl-scale', dest='orthocorls', default=1e-05, type=float,
help='scale factor of ortho_cor_loss (default: 1e-05)')
parser.add_argument('--groupcorl', '--groupcor-loss', dest='groupcor_loss', action='store_true',
help='group correlation loss?')
parser.add_argument('--groupcorls', '--groupcorl-scale', dest='groupcorls', default=1e-05, type=float,
help='scale factor of groupcor_loss (default: 1e-05)')
parser.add_argument('--groupcorn', '--groupcor-num', dest='groupcor_num', default=1, type=int,
help='the number of kernel sets for groups of groupcor_loss (if this number is n, group size is n*3) '
'(available number: 1,2,4,8,16) (default: 1)')
parser.add_argument('-warm', '--warmup-epoch', dest='warmup_epoch', default=0, type=int, metavar='N',
help='number of warmup epochs for applying the V2 method (default: 0)')
parser.add_argument('--w-anal', '--weight-analysis', dest='w_anal', action='store_true',
help='weight analysis in find_similar_kernel.py')
# for quantization
parser.add_argument('-Q', '--quant', dest='quant', action='store_true',
help='use quantization?')
parser.add_argument('--qb', '--quant-bit', default=8, type=int, metavar='N', dest='quant_bit',
help='number of bits for quantization (default: 8)')
parser.add_argument('--qba', '--quant_bit_a', default=8, type=int, metavar='N', dest='quant_bit_a',
help='number of bits for quantizing alphas (default: 8)')
parser.add_argument('--qbb', '--quant_bit_b', default=8, type=int, metavar='N', dest='quant_bit_b',
help='number of bits for quantizing betas (default: 8)')
# for analysis
parser.add_argument('--chk-save', dest='chk_save', action='store_true',
help='save indices and kernel weight for check index variation at fine-tuning')
parser.add_argument('--chk-num', default=0, type=int, metavar='N', dest='chk_num',
help='number of channel number for check index variation at fine-tuning')
cfg = parser.parse_args()
cfg.gpuids = list(map(int, cfg.gpuids))
return cfg