-
Notifications
You must be signed in to change notification settings - Fork 0
/
config.py
67 lines (52 loc) · 3.24 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
import argparse
import datetime
import os
class ArgParser(argparse.ArgumentParser):
def load_arguments(self):
self.add_argument('-T', '--tag', type=str, default='Normal exp P/N with Unc',)
self.add_argument('-L', '--learning-rate', type=float, default=1e-4)
self.add_argument('-E', '--epoch', type=int, default=5)
self.add_argument('-N', '--device', type=str, default='0')
self.add_argument('-LD','--lr-decay', type=float, default=0.001)
self.add_argument('-LDS', '--lr-decay-size', type=int, default=30)
# self.add_argument('-M', '--momentum', type=float, default=0.9,
# help='Momentum value for SGD optimizer.')
self.add_argument('-WD', '--weight-decay', type=float, default=1e-3,
help="Decay weight for optimizer.")
self.add_argument('-RT', '--restart-step', type=int, default=50,
help='Restart step for cosine annealing warm restarts.')
self.add_argument('-BM', '--BN-momentum', type=float, default=0.9,
help="Momentum for batch normalize.")
self.add_argument('-DP', '--dropout', type=float, default=0.1)
self.add_argument('-O', '--output', type=str, default='outputs')
self.add_argument('-DS', '--source-dataset', type=str, default='gcc')
self.add_argument('-DT', '--target-dataset', type=str, default='ucf')
self.add_argument('-TS', '--source-dataset-type', type=str, default='image')
self.add_argument('-TT', '--target-dataset-type', type=str, default='image')
self.add_argument('-TR', '--training-ratio', type=float, default=0.5,
help="Training data ratio, test ratio set as 0.1")
self.add_argument('-B', '--batch-size', type=int, default=8)
self.add_argument('-TSS', '--training-scale-s', type=int, default=500,
help="mbm: 1000, dcc: 500,\
adi: 100, vgg: 100, 'mbc: 1000")
self.add_argument('-TST', '--training-scale-t', type=int, default=500,
help="mbm: 1000, dcc: 500,\
adi: 100, vgg: 100, 'mbc: 1000")
self.add_argument('-RS', '--image-resize', type=int, default=512)
self.add_argument('-P', '--patch-size', type=tuple, default=256,
help="Cropping size of image.")
self.add_argument('-WS', '--warm-start', type=int, default=0,
help="Epochs only train regressor on source domain.")
self.add_argument('-MS', '--memory-saving', type=bool, default=True)
class Constants:
ROOT_PATH = '/home/zhuonan/code/PNUDA' # Baseline directory path
MODEL_NAME = 'UDA' # Model directory path
DATA_FOLDER = os.path.join(ROOT_PATH, 'data')
OUTPUT_FOLDER = os.path.join(ROOT_PATH, 'output')
LOG_FOLDER = os.path.join(ROOT_PATH, 'log')
LOG_NAME = datetime.datetime.now()
TARGET_TRAIN_FILELIST = ''
TARGET_VALID_FILELIST = ''
DATASET = {'vgg': 'VGG', 'mbm':'MBM', 'adi':'ADI', 'mnist':'MNIST',
'mnist_m':'MNIST_moving', 'dcc':'DCC', 'gcc':'GCC', 'ucf':'UCF'}
CFG = [[32, 'R', 'M', 64, 'R', 'M', 128, 'R', 'M', 512, 'R'], [128, 'R', 'U', 64, 'R', 'U', '32', 'R', 'U', 1, 'R']]