-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathmain.py
172 lines (129 loc) · 6.83 KB
/
main.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
"""
Author: Haoran Chen
Date: 2022.08.15
"""
import argparse
import torch
from clip import clip
import os
from torch import nn
from model import Custom_Clip, PromptGenerator
from train_prompt import train_Prompt
from train_msf import train_MSF
from dataloader import load_pseudo_label_data, load_data
import numpy as np
torch.manual_seed(1)
np.random.seed(1)
torch.cuda.manual_seed(1)
torch.cuda.manual_seed_all(1)
torch.backends.cudnn.deterministic = True
def arg_parse():
parser = argparse.ArgumentParser('Training and Evaluation Script', add_help=False)
# for config
parser.add_argument('--file_root', type=str, default=r'/vhome/chenhaoran/hrchen/MPA/',
help='model output path')
parser.add_argument('--data_root', type=str, default=r'/share/test/hrchen/', help='data file path')
parser.add_argument('--backbone', type=str, default='RN101', help='')
parser.add_argument('--dataset', type=str, default='ImageCLEF', help='')
parser.add_argument('--device', type=str, default='cuda', help='')
# for dataloader
parser.add_argument('--batch_size', type=int, default=32, help='')
parser.add_argument('--num_workers', type=int, default=8, help='')
parser.add_argument('--pin_memory', type=bool, default=True, help='')
parser.add_argument('--threshold', type=float, default=0.4, help='threshold tau for generating pseudo labels')
# for prompt settings
parser.add_argument('--M1', type=int, default=12, help='number of classification tokens')
parser.add_argument('--M2', type=int, default=12, help='number of domain tokens')
# for encoder settings
parser.add_argument('--mid_dim', type=int, default=384, help='dimension for the first feed forward layer')
parser.add_argument('--out_dim', type=int, default=250, help='dimension for the intrinsic subspace')
# for training settings
parser.add_argument('--prompt_iteration', type=int, default=5000, help='')
parser.add_argument('--msf_iteration', type=int, default=5000, help='')
parser.add_argument('--prompt_learning_rate', type=float, default=0.01, help='')
parser.add_argument('--prompt_momentum', type=float, default=0.9, help='')
parser.add_argument('--prompt_weight_decay', type=float, default=0.0001, help='')
parser.add_argument('--msf_learning_rate', type=float, default=0.001, help='')
parser.add_argument('--msf_alpha', type=int, default=500, help='')
parser.add_argument('--output_folder', type=str, default='', help='')
parser.add_argument('--n_cls', type=int, default=0, help='number of classes in dataset')
parser.add_argument('--AE_domain', type=bool, default=False, help='number of classes in dataset')
return parser
def args_update(args):
if args.dataset == 'ImageCLEF':
args.backbone = 'RN50'
args.out_dim = 150
args.prompt_iteration = 400
args.msf_iteration = 250
if args.dataset == 'DomainNet':
args.backbone = 'RN101'
args.prompt_iteration = 4000
args.msf_iteration = 2000
args.out_dim = 300
if args.dataset == 'OfficeHome':
args.backbone = 'RN50'
args.out_dim = 150
args.prompt_iteration = 1000
args.msf_iteration = 500
def train(domain_list, classnames, clip_model, preprocess, args):
custom_clip_model = Custom_Clip(clip_model)
custom_clip_model = nn.DataParallel(custom_clip_model)
custom_clip_model = custom_clip_model.module
for name, param in custom_clip_model.named_parameters():
param.requires_grad_(False)
for target_name in domain_list:
source_name_list = domain_list.copy()
source_name_list.remove(target_name)
if not os.path.exists(os.path.join(args.output_folder, target_name)):
os.makedirs(os.path.join(args.output_folder, target_name))
target_path = os.path.join(args.data_root, args.dataset, target_name)
target_train_loader = load_pseudo_label_data(target_name, target_path, preprocess, clip_model, args)
target_test_loader = load_data(target_path, preprocess, args)
prompt_name = []
for source_name in domain_list:
if source_name != target_name:
name = source_name + '2' + target_name + '.pkl'
prompt_name.append(name)
if os.path.exists(args.output_folder + '/' + target_name + '/' + name):
continue
source_path = os.path.join(args.data_root, args.dataset, source_name)
source_train_loader = load_data(source_path, preprocess, args)
print("Start training {} to {} prompt".format(source_name, target_name))
train_Prompt(target_train_loader, target_test_loader, source_train_loader, classnames, clip_model,
custom_clip_model, source_name, target_name, args)
print("===========================================================================================")
prompt_cls_list = []
prompt_domain_list = []
for i in range(len(prompt_name)):
name = prompt_name[i]
source_name = source_name_list[i]
prompt_learner = PromptGenerator(classnames, clip_model, source_name, target_name, args)
prompt_learner.load_state_dict(torch.load(args.output_folder + '/' + target_name + '/' + name))
ctx_cls = prompt_learner.ctx_cls.float()
ctx_source = prompt_learner.ctx_source.float()
ctx_target = prompt_learner.ctx_target.float()
prompt_cls_list.append(ctx_cls)
prompt_domain_list.append(ctx_target)
print("Start aligning {} prompts".format(target_name))
train_MSF(target_name, target_train_loader, target_test_loader, prompt_cls_list, prompt_domain_list, custom_clip_model, clip_model, classnames, args)
print("===========================================================================================")
def main(args):
args_update(args)
model_path = args.file_root + args.backbone + '.pt'
model, preprocess = clip.load(args.backbone, device=args.device, model_path=model_path)
domain_list = os.listdir(args.data_root + args.dataset)
domain_list = [x for x in domain_list if '.txt' not in x]
classnames_path = os.path.join(args.data_root, args.dataset, domain_list[0])
classnames = os.listdir(classnames_path)
n_cls = len(classnames)
classnames.sort()
args.output_folder = os.path.join(args.file_root, args.dataset, args.backbone, 'MPA_FINAL')
if not os.path.exists(args.output_folder):
os.makedirs(args.output_folder)
args.n_cls = n_cls
print(args)
train(domain_list, classnames, model, preprocess, args)
if __name__ == '__main__':
parser = argparse.ArgumentParser('Training and Evaluation Script', parents=[arg_parse()])
args = parser.parse_args()
main(args)