-
Notifications
You must be signed in to change notification settings - Fork 5
/
Copy pathtrain.py
248 lines (209 loc) · 11.5 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
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
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
import os.path as osp
import torch
import torch.nn.functional as F
from torch.autograd import Variable
import datasets
import settings
from metric import compress_wiki, compress, calculate_top_map, calculate_map, p_topK
from models import ImgNet, TxtNet
torch.manual_seed(1)
torch.cuda.manual_seed_all(1)
class Session:
def __init__(self):
self.logger = settings.logger
torch.cuda.set_device(settings.GPU_ID)
if settings.DATASET == "WIKI":
self.train_dataset = datasets.WIKI(root=settings.DATA_DIR, train=True,
transform=datasets.wiki_train_transform)
self.test_dataset = datasets.WIKI(root=settings.DATA_DIR, train=False,
transform=datasets.wiki_test_transform)
self.database_dataset = datasets.WIKI(root=settings.DATA_DIR, train=True,
transform=datasets.wiki_test_transform)
if settings.DATASET == "MIRFlickr":
self.train_dataset = datasets.MIRFlickr(train=True, transform=datasets.mir_train_transform)
self.test_dataset = datasets.MIRFlickr(train=False, database=False, transform=datasets.mir_test_transform)
self.database_dataset = datasets.MIRFlickr(train=False, database=True,
transform=datasets.mir_test_transform)
if settings.DATASET == "NUSWIDE":
self.train_dataset = datasets.NUSWIDE(train=True, transform=datasets.nus_train_transform)
self.test_dataset = datasets.NUSWIDE(train=False, database=False, transform=datasets.nus_test_transform)
self.database_dataset = datasets.NUSWIDE(train=False, database=True, transform=datasets.nus_test_transform)
# Data Loader (Input Pipeline)
self.train_loader = torch.utils.data.DataLoader(dataset=self.train_dataset,
batch_size=settings.BATCH_SIZE,
shuffle=True,
num_workers=settings.NUM_WORKERS,
drop_last=True)
self.test_loader = torch.utils.data.DataLoader(dataset=self.test_dataset,
batch_size=settings.BATCH_SIZE,
shuffle=False,
num_workers=settings.NUM_WORKERS)
self.database_loader = torch.utils.data.DataLoader(dataset=self.database_dataset,
batch_size=settings.BATCH_SIZE,
shuffle=False,
num_workers=settings.NUM_WORKERS)
txt_feat_len = datasets.txt_feat_len
self.CodeNet_I = ImgNet(code_len=settings.CODE_LEN, txt_feat_len=txt_feat_len)
self.FeatNet_I = ImgNet(code_len=settings.CODE_LEN, txt_feat_len=txt_feat_len)
self.CodeNet_T = TxtNet(code_len=settings.CODE_LEN, txt_feat_len=txt_feat_len)
if settings.DATASET == "WIKI":
self.opt_I = torch.optim.SGD(self.CodeNet_I.parameters(), lr=settings.LR_IMG, momentum=settings.MOMENTUM,
weight_decay=settings.WEIGHT_DECAY, nesterov=True)
if settings.DATASET == "MIRFlickr" or settings.DATASET == "NUSWIDE":
self.opt_I = torch.optim.SGD(self.CodeNet_I.parameters(), lr=settings.LR_IMG, momentum=settings.MOMENTUM,
weight_decay=settings.WEIGHT_DECAY, nesterov=True)
self.opt_T = torch.optim.SGD(self.CodeNet_T.parameters(), lr=settings.LR_TXT, momentum=settings.MOMENTUM,
weight_decay=settings.WEIGHT_DECAY, nesterov=True)
self.best = 0
def train(self, epoch):
self.FeatNet_I.cuda().eval()
self.CodeNet_I.cuda().train()
self.CodeNet_T.cuda().train()
self.CodeNet_I.set_alpha(epoch)
self.CodeNet_T.set_alpha(epoch)
self.logger.info('Epoch [%d/%d], alpha for ImgNet: %.3f, alpha for TxtNet: %.3f' % (
epoch + 1, settings.NUM_EPOCH, self.CodeNet_I.alpha, self.CodeNet_T.alpha))
for idx, (img, txt, labels, _) in enumerate(self.train_loader):
batch_size = img.size(0)
img = Variable(img.cuda())
txt = Variable(torch.FloatTensor(txt.numpy()).cuda())
self.opt_I.zero_grad()
self.opt_T.zero_grad()
(_, F_I), _, _, _ = self.FeatNet_I(img)
F_T = txt
_, hid_I, code_I, decoded_t = self.CodeNet_I(img)
_, hid_T, code_T, decoded_i = self.CodeNet_T(txt)
F_I = F.normalize(F_I)
S_I = F_I.mm(F_I.t())
S_I = S_I * 2 - 1
F_T = F.normalize(F_T)
S_T = F_T.mm(F_T.t())
S_T = S_T * 2 - 1
B_I = F.normalize(code_I)
B_T = F.normalize(code_T)
BI_BI = B_I.mm(B_I.t())
BT_BT = B_T.mm(B_T.t())
BI_BT = B_I.mm(B_T.t())
S_tilde = settings.ALPHA * S_I + (1 - settings.ALPHA) * S_T
S = settings.K * S_tilde
loss1 = F.mse_loss(BT_BT, S)
loss2 = F.mse_loss(BI_BT, S)
loss3 = F.mse_loss(BI_BI, S)
loss31 = F.mse_loss(BI_BI, settings.K * S_I)
loss32 = F.mse_loss(BT_BT, settings.K * S_T)
diagonal = BI_BT.diagonal()
all_1 = torch.rand((batch_size)).fill_(1).cuda()
loss4 = F.mse_loss(diagonal, settings.K * all_1)
loss5 = F.mse_loss(decoded_i, F_I)
loss6 = F.mse_loss(decoded_t, F_T)
loss7 = F.mse_loss(BI_BT, BI_BT.t())
loss = 1 * loss1 + 1 * loss2 + 1 * loss3 + 1 * loss4 + 1 * loss5 + 1 * loss6 + 2 * loss7 + settings.ETA * (
loss31 + loss32)
# MIRFlickr
# 16 bit 863 846; 860 849; 856 841
# 32 bit 877 860; 867 858; 873 856
# 64 bit 889 883; 886 888; 895 881
# 128bit 903 882; 897 881; 907 877; 901 885
# Wiki
# 128bit 435 662;432 661;434 667; 433，663;
# 64 bit 440 658;438 660;433 660;
# 32 bit 430 650;422 658;420 665;
# 16 bit 394 617;416 644;416 639;
# NUS-WIDE
# 128bit
# 64 bit
# 32 bit
# 16 bit
loss.backward()
self.opt_I.step()
self.opt_T.step()
if (idx + 1) % (len(self.train_dataset) // settings.BATCH_SIZE / settings.EPOCH_INTERVAL) == 0:
self.logger.info(
'Epoch [%d/%d], Iter [%d/%d] '
'Loss1: %.4f Loss2: %.4f Loss3: %.4f '
'Loss4: %.4f '
'Loss5: %.4f Loss6: %.4f '
'Loss7: %.4f '
'Total Loss: %.4f'
% (
epoch + 1, settings.NUM_EPOCH, idx + 1,
len(self.train_dataset) // settings.BATCH_SIZE,
loss1.item(), loss2.item(), loss3.item(),
loss4.item(),
loss5.item(), loss6.item(),
loss7.item(),
loss.item()))
def eval(self, step=0, last=False):
# Change model to 'eval' mode (BN uses moving mean/var).
self.CodeNet_I.eval().cuda()
self.CodeNet_T.eval().cuda()
if settings.DATASET == "WIKI":
re_BI, re_BT, re_L, qu_BI, qu_BT, qu_L = compress_wiki(self.database_loader, self.test_loader,
self.CodeNet_I, self.CodeNet_T,
self.database_dataset, self.test_dataset)
K = [1, 200, 400, 500, 1000, 1500, 2000]
if settings.DATASET == "MIRFlickr" or settings.DATASET == "NUSWIDE":
re_BI, re_BT, re_L, qu_BI, qu_BT, qu_L = compress(self.database_loader, self.test_loader, self.CodeNet_I,
self.CodeNet_T, self.database_dataset, self.test_dataset)
K = [1, 200, 500, 1000, 1500, 2000, 2500, 3000, 3500, 4000, 4500, 5000]
if settings.EVAL:
MAP_I2T = calculate_map(qu_B=qu_BI, re_B=re_BT, qu_L=qu_L, re_L=re_L)
MAP_T2I = calculate_map(qu_B=qu_BT, re_B=re_BI, qu_L=qu_L, re_L=re_L)
self.logger.info('--------------------Evaluation: Calculate top MAP-------------------')
self.logger.info('MAP of Image to Text: %.3f, MAP of Text to Image: %.3f' % (MAP_I2T, MAP_T2I))
self.logger.info('--------------------------------------------------------------------')
retI2T = p_topK(qu_BI, re_BT, qu_L, re_L, K)
retT2I = p_topK(qu_BT, re_BI, qu_L, re_L, K)
self.logger.info(retI2T)
self.logger.info(retT2I)
MAP_I2T = calculate_top_map(qu_B=qu_BI, re_B=re_BT, qu_L=qu_L, re_L=re_L, topk=50)
MAP_T2I = calculate_top_map(qu_B=qu_BT, re_B=re_BI, qu_L=qu_L, re_L=re_L, topk=50)
self.logger.info('--------------------Evaluation: Calculate top MAP-------------------')
self.logger.info('MAP of Image to Text: %.3f, MAP of Text to Image: %.3f' % (MAP_I2T, MAP_T2I))
self.logger.info('--------------------------------------------------------------------')
if MAP_I2T + MAP_T2I > self.best and not settings.EVAL:
self.save_checkpoints(step=step, best=True)
self.best = MAP_T2I + MAP_I2T
self.logger.info("#########is best:%.3f #########" % self.best)
def save_checkpoints(self, step, file_name='%s_%d_bit_latest.pth' % (settings.DATASET, settings.CODE_LEN),
best=False):
if best:
file_name = '%s_%d_bit_best_epoch.pth' % (settings.DATASET, settings.CODE_LEN)
ckp_path = osp.join(settings.MODEL_DIR, file_name)
obj = {
'ImgNet': self.CodeNet_I.state_dict(),
'TxtNet': self.CodeNet_T.state_dict(),
'step': step,
}
torch.save(obj, ckp_path)
self.logger.info('**********Save the trained model successfully.**********')
def load_checkpoints(self, file_name='%s_%d_bit_best_epoch.pth' % (settings.DATASET, settings.CODE_LEN)):
ckp_path = osp.join(settings.MODEL_DIR, file_name)
try:
obj = torch.load(ckp_path, map_location=lambda storage, loc: storage.cuda())
self.logger.info('**************** Load checkpoint %s ****************' % ckp_path)
except IOError:
self.logger.error('********** No checkpoint %s!*********' % ckp_path)
return
self.CodeNet_I.load_state_dict(obj['ImgNet'])
self.CodeNet_T.load_state_dict(obj['TxtNet'])
self.logger.info('********** The loaded model has been trained for %d epochs.*********' % obj['step'])
def main():
sess = Session()
if settings.EVAL == True:
sess.load_checkpoints()
sess.eval()
else:
for epoch in range(settings.NUM_EPOCH):
# train the Model
sess.train(epoch)
# eval the Model
if (epoch + 1) % settings.EVAL_INTERVAL == 0:
sess.eval(step=epoch + 1)
# save the model
settings.EVAL = True
sess.logger.info('---------------------------Test------------------------')
sess.load_checkpoints()
sess.eval()
if __name__ == '__main__':
main()