-
Notifications
You must be signed in to change notification settings - Fork 4
/
Copy pathtrain.py
135 lines (111 loc) · 6.02 KB
/
train.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
import argparse
import importlib
from utils import *
import torch
MODEL_DIR=None
DATA_DIR = './local_datasets/'
PROJECT='base'
def get_command_line_parser():
parser = argparse.ArgumentParser()
# about dataset and network
parser.add_argument('-project', type=str, default=PROJECT)
parser.add_argument('-dataset', type=str, default='cifar100',
choices=['mini_imagenet', 'cub200', 'cifar100'])
parser.add_argument('-dataroot', type=str, default=DATA_DIR)
parser.add_argument('-out', type=str, default=None)
# about pre-training
# parser.add_argument('-epochs_base', type=int, default=200)
parser.add_argument('-epochs_base', type=int, default=50)
# parser.add_argument('-epochs_new', type=int, default=100)
parser.add_argument('-epochs_new', type=int, default=20)
parser.add_argument('-lr_base', type=float, default=0.1)
parser.add_argument('-lr_new', type=float, default=2e-4)
parser.add_argument('-schedule', type=str, default='Step',
choices=['Step', 'Milestone', 'Cosine'])
parser.add_argument('-milestones', nargs='+', type=int, default=[60, 70])
parser.add_argument('-step', type=int, default=80)
parser.add_argument('-decay', type=float, default=0.0005)
parser.add_argument('-momentum', type=float, default=0.9)
parser.add_argument('-gamma', type=float, default=0.1)
parser.add_argument('-temperature', type=int, default=16)
parser.add_argument('-not_data_init', action='store_true', help='using average data embedding to init or not')
parser.add_argument('-batch_size_base', type=int, default=128)
parser.add_argument('-batch_size_new', type=int, default=0, help='set 0 will use all the availiable training image for new')
parser.add_argument('-test_batch_size', type=int, default=128)
parser.add_argument('-base_mode', type=str, default='ft_cos',
choices=['ft_dot', 'ft_cos']) # ft_dot means using linear classifier, ft_cos means using cosine classifier
parser.add_argument('-new_mode', type=str, default='avg_cos',
choices=['ft_dot', 'ft_cos', 'avg_cos', 'ft_comb', 'ft_euc']) # ft_dot means using linear classifier, ft_cos means using cosine classifier, avg_cos means using average data embedding and cosine classifier
# for episode learning
parser.add_argument('-train_episode', type=int, default=50)
parser.add_argument('-episode_shot', type=int, default=1)
parser.add_argument('-episode_way', type=int, default=15)
parser.add_argument('-episode_query', type=int, default=15)
# for cec
parser.add_argument('-lrg', type=float, default=0.1) #lr for graph attention network
parser.add_argument('-low_shot', type=int, default=1)
parser.add_argument('-low_way', type=int, default=15)
parser.add_argument('-start_session', type=int, default=1)
parser.add_argument('-model_dir', type=str, default=MODEL_DIR, help='loading model parameter from a specific dir')
parser.add_argument('-set_no_val', action='store_true', help='set validation using test set or no validation')
# about training
parser.add_argument('-gpu', default='2, 3')
parser.add_argument('-num_workers', type=int, default=4)
parser.add_argument('-seed', type=int, default=-1)
# parser.add_argument('-seeds', nargs='+', help='<Required> Set flag', required=True, default=1)
parser.add_argument('-debug', action='store_true')
parser.add_argument('-rotation', action='store_true')
parser.add_argument('-fraction_to_keep', type=float, default=0.1)
parser.add_argument('-vit', action='store_true')
parser.add_argument('-baseline', action='store_true')
parser.add_argument('-clip', action='store_true')
parser.add_argument('-ED', action='store_true')
parser.add_argument('-ED_hp', type=float, default=0.1)
parser.add_argument('-LT', action='store_true')
parser.add_argument('-WC', action='store_true')
parser.add_argument('-MP', action='store_true')
parser.add_argument('-SKD', action='store_true')
parser.add_argument('-l2p', action='store_true')
parser.add_argument('-dp', action='store_true')
parser.add_argument('-prefix', action='store_true')
parser.add_argument('-pret_clip', action='store_true')
parser.add_argument('-comp_out', type=int, default=1.)
# parser.add_argument('-scratch', action='store_true')
parser.add_argument('-scratch', action='store_true')
parser.add_argument('-taskblock', type=int, default=2)
parser.add_argument('-ft', action='store_true')
parser.add_argument('-lp', action='store_true')
parser.add_argument('-PKT_tune_way', type=int, default=1)
return parser
if __name__ == '__main__':
torch.set_num_threads(2)
devices = [d for d in range(torch.cuda.device_count())]
device_names = [torch.cuda.get_device_name(d) for d in devices]
print('GPU COUNT',torch.cuda.device_count())
print("Devices:",devices)
print("Device Name:",device_names)
parser = get_command_line_parser()
args = parser.parse_args()
# For incremental learning, get same random seed that has been used during pretraining step
# if args.model_dir is not None:
# args.seed = int(args.model_dir.split('/')[-2][-9])
# for seed in args.seed:
# args.seed = seed
set_seed(args.seed)
pprint(vars(args))
args.num_gpu = set_gpu(args)
if args.vit:
if args.baseline:
trainer = importlib.import_module('models.%s.ViT_fscil_Baseline_trainer' % (args.project)).ViT_Baseline_FSCILTrainer(args)
#* L2P / DP
elif args.l2p:
pass
elif args.dp:
pass
else:
trainer = importlib.import_module('models.%s.ViT_fscil_trainer' % (args.project)).ViT_FSCILTrainer(args)
elif args.clip:
trainer = importlib.import_module('models.%s.CLIP_fscil_trainer' % (args.project)).CLIP_FSCILTrainer(args)
else:
trainer = importlib.import_module('models.%s.fscil_trainer' % (args.project)).FSCILTrainer(args)
trainer.train()