-
Notifications
You must be signed in to change notification settings - Fork 5
/
Copy pathcustom_config.py
95 lines (68 loc) · 2.2 KB
/
custom_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
from yacs.config import CfgNode as CN
_C = CN()
_C.OUTPUT_DIR = "output/meta_color/"
_C.NORM_FLAG = True
_C.SEED = 666
_C.DATA = CN()
_C.DATA.DATASET='ZipDataset'
_C.DATA.SUB_DIR = "EXT0.0"
_C.DATA.TRAIN_SRC_REAL_1 = 'data/data_list/CASIA-ALL-REAL.csv'
_C.DATA.TRAIN_SRC_FAKE_1 = 'data/data_list/CASIA-ALL-FAKE.csv'
_C.DATA.TRAIN_SRC_REAL_2 = 'data/data_list/MSU-MFSD-REAL.csv'
_C.DATA.TRAIN_SRC_FAKE_2 = 'data/data_list/MSU-MFSD-FAKE.csv'
_C.DATA.TRAIN_SRC_REAL_3 = 'data/data_list/REPLAY-ALL-REAL.csv'
_C.DATA.TRAIN_SRC_FAKE_3 = 'data/data_list/REPLAY-ALL-FAKE.csv'
_C.DATA.TARGET_DATA = 'data/data_list/OULU-NPU.csv'
_C.DATA.BATCH_SIZE = 4
_C.DATA.TRAIN_NF = 1000
_C.DATA.VAL_NF = 2
_C.DATA.TEST_NF = 2
_C.MODEL = CN()
_C.TRAIN = CN()
_C.TRAIN.INIT_LR = 0.01
_C.TRAIN.LR_EPOCH_1 = 0
_C.TRAIN.LR_EPOCH_2 = 150
_C.TRAIN.MOMENTUM = 0.9
_C.TRAIN.WEIGHT_DECAY = 0.0 # 5e-4
_C.TRAIN.WEIGHT_DECAY_T = 0.0 # ColorNet for TRANSFORMER
_C.TRAIN.MAX_ITER = 1000000
_C.TRAIN.META_TRAIN_SIZE = 2
_C.TRAIN.ITER_PER_EPOCH = 100
_C.TRAIN.META_PRE_TRAIN = True
_C.TRAIN.DROPOUT = 0.0
_C.TRAIN.SYNC_TRAINING = False
_C.TRAIN.IMAGENET_PRETRAIN = True
_C.TRAIN.AUG = CN(new_allowed=True)
_C.TRAIN.AUG.RandomHorizontalFlip = CN(new_allowed=True)
_C.TRAIN.AUG.RandomHorizontalFlip.ENABLE = True
_C.TRAIN.AUG.RandomHorizontalFlip.p = 0.5
# TODO
_C.TRAIN.W_depth = 10
_C.TRAIN.W_metatest = 1
_C.TRAIN.META_LEARNING_RATE = 0.001
_C.TRAIN.BETAS = [0.9, 0.999]
_C.TRAIN.META_TEST_FREQ = 1
_C.TRAIN.VAL_FREQ = 1
_C.TRAIN.NUM_FRAMES = 1000
_C.TRAIN.INNER_LOOPS = 1
_C.TRAIN.RETRAIN_FROM_SCATCH = True
_C.TRAIN.OPTIM = 'SGD' # Adam
# ['None', 'CosineAnnealingLR']
_C.TRAIN.LR_SCHEDULER = CN()
_C.TRAIN.LR_SCHEDULER.NAME = ''
_C.TRAIN.LR_SCHEDULER.CosineAnnealingLR = CN()
_C.TRAIN.LR_SCHEDULER.CosineAnnealingLR.T_max = 400
_C.TRAIN.LR_SCHEDULER.CosineAnnealingLR.eta_min = 0.000001
_C.TRAIN.LR_SCHEDULER.CosineAnnealingLR.last_epoch = -1
_C.MODEL = CN()
_C.MODEL.IN_CHANNELS = 3
_C.MODEL.MEAN_STD_NORMAL = True
_C.MODEL.CHANNELS = CN()
_C.MODEL.CHANNELS.RGB = True
_C.MODEL.CHANNELS.HSV = False
_C.MODEL.CHANNELS.YCRCB = False
_C.MODEL.CHANNELS.YUV = False
_C.MODEL.CHANNELS.LAB = False
_C.MODEL.CHANNELS.XYZ = False
_C.TEST = CN()
_C.TEST.NUM_FRAMES = 2