-
Notifications
You must be signed in to change notification settings - Fork 0
/
train_UT-KD.py
477 lines (400 loc) · 24.5 KB
/
train_UT-KD.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
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
import argparse
import torch
import torch.nn as nn
from torch.utils import data
import numpy as np
from torch.autograd import Variable
import torch.optim as optim
import torch.nn.functional as F
import os
import time
from utils.tools import *
from dataset.cityscapes_dataset import cityscapesDataSet
from dataset.idd_dataset import IDDDataSet
from dataset.mapillary_dataset import MapDataSet
from model.Networks import DeeplabMulti,Discriminator
import torch.backends.cudnn as cudnn
import mask_gen
IMG_MEAN = np.array((104.00698793, 116.66876762, 122.67891434), dtype=np.float32)
def get_arguments():
parser = argparse.ArgumentParser()
#dataset
parser.add_argument("--data_ID", type=int, default=1)
parser.add_argument("--input_size_tgt", type=str, default='1024,512',
help="width and height of input target images.")
parser.add_argument("--num_classes", type=int, default=7,
help="number of classes.")
#network
parser.add_argument("--batch_size", type=int, default=2,
help="number of images in each batch.")
parser.add_argument("--num_workers", type=int, default=0,
help="number of workers for multithread dataloading.")
parser.add_argument("--learning_rate", type=float, default=2.5e-5,
help="base learning rate.")
parser.add_argument("--learning_rate_D", type=float, default=1e-5,
help="discriminator learning rate.")
parser.add_argument("--momentum", type=float, default=0.9,
help="momentum.")
parser.add_argument("--num_steps", type=int, default=50000,
help="Number of training steps.")
parser.add_argument("--num_steps_stop", type=int, default=50000,
help="Number of training steps for early stopping.")
parser.add_argument("--restore_from", type=str, default='./pretrained_model/GTA_pretrain_batch_50000_miou_67.pth',
help="pretrained model from GTA-5.")
parser.add_argument("--weight_decay", type=float, default=5e-5,
help="regularisation parameter for L2-loss.")
#hyperparameters
parser.add_argument("--teacher_alpha", type=float, default=0.999,
help="teacher alpha in EMA.")
parser.add_argument("--con_weight", type=float, default=100,
help="consistency regularization weight.")
parser.add_argument("--out_weight", type=float, default=0.1,
help="adversarial training weight.")
#result
parser.add_argument("--snapshot_dir", type=str, default='./UT-KD/',
help="where to save snapshots of the model.")
return parser.parse_args()
def main():
args = get_arguments()
w, h = map(int, args.input_size_tgt.split(','))
input_size_tgt = (w, h)
if args.data_ID == 1:
# synthetic2real_GTA2City2IDD
seg_restore_from = './pretrained_model/MT-KD/synthetic2real_GTA2City/Seg_batch_12000_miou_7631.pth'
D1_restore_from = './pretrained_model/MT-KD/synthetic2real_GTA2City/D_net1_batch_12000.pth'
D2_restore_from = './pretrained_model/MT-KD/synthetic2real_GTA2City/D_net2_batch_12000.pth'
data_dir_target = '/iarai/home/yonghao.xu/Data/SegmentationData/IDD/IDD_Segmentation/'
data_list_val = './dataset/IDD_val.txt'
data_list_target = './dataset/IDD_train.txt'
input_size_test = (1920,1080)
snapshot_dir = args.snapshot_dir+'synthetic2real_GTA2City2IDD/UTKD_out_weight_'+str(args.out_weight)+'_con_weight_'+str(args.con_weight)+'_time'+time.strftime('%m%d_%H%M', time.localtime(time.time()))+'/'
tgt_loader = data.DataLoader(
IDDDataSet(data_dir_target, data_list_target, max_iters=args.num_steps_stop*args.batch_size,
crop_size=input_size_tgt,
scale=False, mirror=False, mean=IMG_MEAN,
set='train'),
batch_size=args.batch_size, shuffle=True, num_workers=args.num_workers,
pin_memory=True)
val_loader = data.DataLoader(
IDDDataSet(data_dir_target, data_list_val, max_iters=None,
crop_size=input_size_tgt,
scale=False, mirror=False, mean=IMG_MEAN,
set='val'),
batch_size=args.batch_size, shuffle=False, num_workers=args.num_workers,
pin_memory=True)
elif args.data_ID == 2:
# synthetic2real_GTA2City2Mapillary
seg_restore_from = './pretrained_model/MT-KD/synthetic2real_GTA2City/Seg_batch_12000_miou_7631.pth'
D1_restore_from = './pretrained_model/MT-KD/synthetic2real_GTA2City/D_net1_batch_12000.pth'
D2_restore_from = './pretrained_model/MT-KD/synthetic2real_GTA2City/D_net2_batch_12000.pth'
data_dir_target = '/iarai/home/yonghao.xu/Data/SegmentationData/MapillaryVistas/'
data_list_val = './dataset/Mapillary_val.txt'
data_list_target = './dataset/Mapillary_train.txt'
input_size_test = (1920,1080)
snapshot_dir = args.snapshot_dir+'synthetic2real_GTA2City2Mapillary/UTKD_out_weight_'+str(args.out_weight)+'_con_weight_'+str(args.con_weight)+'_time'+time.strftime('%m%d_%H%M', time.localtime(time.time()))+'/'
tgt_loader = data.DataLoader(
MapDataSet(data_dir_target, data_list_target, max_iters=args.num_steps_stop*args.batch_size,
crop_size=input_size_tgt,
scale=False, mirror=False, mean=IMG_MEAN,
set='train'),
batch_size=args.batch_size, shuffle=True, num_workers=args.num_workers,
pin_memory=True)
val_loader = data.DataLoader(
MapDataSet(data_dir_target, data_list_val, max_iters=None,
crop_size=input_size_tgt,
scale=False, mirror=False, mean=IMG_MEAN,
set='val'),
batch_size=args.batch_size, shuffle=False, num_workers=args.num_workers,
pin_memory=True)
elif args.data_ID == 3:
# synthetic2real_GTA2IDD2City
seg_restore_from = './pretrained_model/MT-KD/synthetic2real_GTA2IDD/Seg_batch_20500_miou_7145.pth'
D1_restore_from = './pretrained_model/MT-KD/synthetic2real_GTA2IDD/D_net1_batch_20500.pth'
D2_restore_from = './pretrained_model/MT-KD/synthetic2real_GTA2IDD/D_net2_batch_20500.pth'
data_list_target = './dataset/cityscapes_labellist_train.txt'
data_list_val = './dataset/cityscapes_labellist_val.txt'
data_dir_target = '/iarai/home/yonghao.xu/Data/SegmentationData/Cityscapes/'
input_size_test = (2048,1024)
snapshot_dir = args.snapshot_dir+'synthetic2real_GTA2IDD2City/UTKD_out_weight_'+str(args.out_weight)+'_con_weight_'+str(args.con_weight)+'_time'+time.strftime('%m%d_%H%M', time.localtime(time.time()))+'/'
tgt_loader = data.DataLoader(
cityscapesDataSet(data_dir_target, data_list_target, max_iters=args.num_steps_stop*args.batch_size,
crop_size=input_size_tgt,
scale=False, mirror=False, mean=IMG_MEAN,
set='train',num_classes=args.num_classes),
batch_size=args.batch_size, shuffle=True, num_workers=args.num_workers,
pin_memory=True)
val_loader = data.DataLoader(
cityscapesDataSet(data_dir_target, data_list_val, max_iters=None,
crop_size=input_size_tgt,
scale=False, mirror=False, mean=IMG_MEAN,
set='val',num_classes=args.num_classes),
batch_size=args.batch_size, shuffle=False, num_workers=args.num_workers,
pin_memory=True)
elif args.data_ID == 4:
# synthetic2real_GTA2Mapillary2City
seg_restore_from = './pretrained_model/MT-KD/synthetic2real_GTA2Mapillary/Seg_batch_15500_miou_7523.pth'
D1_restore_from = './pretrained_model/MT-KD/synthetic2real_GTA2Mapillary/D_net1_batch_15500.pth'
D2_restore_from = './pretrained_model/MT-KD/synthetic2real_GTA2Mapillary/D_net2_batch_15500.pth'
data_list_target = './dataset/cityscapes_labellist_train.txt'
data_list_val = './dataset/cityscapes_labellist_val.txt'
data_dir_target = '/iarai/home/yonghao.xu/Data/SegmentationData/Cityscapes/'
input_size_test = (2048,1024)
snapshot_dir = args.snapshot_dir+'synthetic2real_GTA2Mapillary2City/UTKD_out_weight_'+str(args.out_weight)+'_con_weight_'+str(args.con_weight)+'_time'+time.strftime('%m%d_%H%M', time.localtime(time.time()))+'/'
tgt_loader = data.DataLoader(
cityscapesDataSet(data_dir_target, data_list_target, max_iters=args.num_steps_stop*args.batch_size,
crop_size=input_size_tgt,
scale=False, mirror=False, mean=IMG_MEAN,
set='train',num_classes=args.num_classes),
batch_size=args.batch_size, shuffle=True, num_workers=args.num_workers,
pin_memory=True)
val_loader = data.DataLoader(
cityscapesDataSet(data_dir_target, data_list_val, max_iters=None,
crop_size=input_size_tgt,
scale=False, mirror=False, mean=IMG_MEAN,
set='val',num_classes=args.num_classes),
batch_size=args.batch_size, shuffle=False, num_workers=args.num_workers,
pin_memory=True)
elif args.data_ID == 5:
# synthetic2real_GTA2IDDMapillary2City
seg_restore_from = './pretrained_model/MT-KD/synthetic2real_GTA2IDDMapillary/Seg_batch_15500_miou_7090_val2_miou_7531.pth'
D1_restore_from = './pretrained_model/MT-KD/synthetic2real_GTA2IDDMapillary/D_net1_batch_15500.pth'
D2_restore_from = './pretrained_model/MT-KD/synthetic2real_GTA2IDDMapillary/D_net2_batch_15500.pth'
data_list_target = './dataset/cityscapes_labellist_train.txt'
data_list_val = './dataset/cityscapes_labellist_val.txt'
data_dir_target = '/iarai/home/yonghao.xu/Data/SegmentationData/Cityscapes/'
input_size_test = (2048,1024)
snapshot_dir = args.snapshot_dir+'synthetic2real_GTA2IDDMapillary2City/UTKD_out_weight_'+str(args.out_weight)+'_con_weight_'+str(args.con_weight)+'_time'+time.strftime('%m%d_%H%M', time.localtime(time.time()))+'/'
tgt_loader = data.DataLoader(
cityscapesDataSet(data_dir_target, data_list_target, max_iters=args.num_steps_stop*args.batch_size,
crop_size=input_size_tgt,
scale=False, mirror=False, mean=IMG_MEAN,
set='train',num_classes=args.num_classes),
batch_size=args.batch_size, shuffle=True, num_workers=args.num_workers,
pin_memory=True)
val_loader = data.DataLoader(
cityscapesDataSet(data_dir_target, data_list_val, max_iters=None,
crop_size=input_size_tgt,
scale=False, mirror=False, mean=IMG_MEAN,
set='val',num_classes=args.num_classes),
batch_size=args.batch_size, shuffle=False, num_workers=args.num_workers,
pin_memory=True)
elif args.data_ID == 6:
# synthetic2real_GTA2CityMapillary2IDD
seg_restore_from = './pretrained_model/MT-KD/synthetic2real_GTA2CityMapillary/Seg_batch_9000_miou_7683_val2_miou_7536.pth'
D1_restore_from = './pretrained_model/MT-KD/synthetic2real_GTA2CityMapillary/D_net1_batch_9000.pth'
D2_restore_from = './pretrained_model/MT-KD/synthetic2real_GTA2CityMapillary/D_net2_batch_9000.pth'
data_dir_target = '/iarai/home/yonghao.xu/Data/SegmentationData/IDD/IDD_Segmentation/'
data_list_val = './dataset/IDD_val.txt'
data_list_target = './dataset/IDD_train.txt'
input_size_test = (1920,1080)
snapshot_dir = args.snapshot_dir+'synthetic2real_GTA2CityMapillary2IDD/UTKD_out_weight_'+str(args.out_weight)+'_con_weight_'+str(args.con_weight)+'_time'+time.strftime('%m%d_%H%M', time.localtime(time.time()))+'/'
tgt_loader = data.DataLoader(
IDDDataSet(data_dir_target, data_list_target, max_iters=args.num_steps_stop*args.batch_size,
crop_size=input_size_tgt,
scale=False, mirror=False, mean=IMG_MEAN,
set='train'),
batch_size=args.batch_size, shuffle=True, num_workers=args.num_workers,
pin_memory=True)
val_loader = data.DataLoader(
IDDDataSet(data_dir_target, data_list_val, max_iters=None,
crop_size=input_size_tgt,
scale=False, mirror=False, mean=IMG_MEAN,
set='val'),
batch_size=args.batch_size, shuffle=False, num_workers=args.num_workers,
pin_memory=True)
elif args.data_ID == 7:
# synthetic2real_GTA2CityIDD2Mapillary
seg_restore_from = './pretrained_model/MT-KD/synthetic2real_GTA2CityIDD/Seg_batch_14000_miou_7654_val2_miou_7122.pth'
D1_restore_from = './pretrained_model/MT-KD/synthetic2real_GTA2CityIDD/D_net1_batch_14000.pth'
D2_restore_from = './pretrained_model/MT-KD/synthetic2real_GTA2CityIDD/D_net2_batch_14000.pth'
data_dir_target = '/iarai/home/yonghao.xu/Data/SegmentationData/MapillaryVistas/'
data_list_val = './dataset/Mapillary_val.txt'
data_list_target = './dataset/Mapillary_train.txt'
input_size_test = (1920,1080)
snapshot_dir = args.snapshot_dir+'synthetic2real_GTA2CityIDD2Mapillary/UTKD_out_weight_'+str(args.out_weight)+'_con_weight_'+str(args.con_weight)+'_time'+time.strftime('%m%d_%H%M', time.localtime(time.time()))+'/'
tgt_loader = data.DataLoader(
MapDataSet(data_dir_target, data_list_target, max_iters=args.num_steps_stop*args.batch_size,
crop_size=input_size_tgt,
scale=False, mirror=False, mean=IMG_MEAN,
set='train'),
batch_size=args.batch_size, shuffle=True, num_workers=args.num_workers,
pin_memory=True)
val_loader = data.DataLoader(
MapDataSet(data_dir_target, data_list_val, max_iters=None,
crop_size=input_size_tgt,
scale=False, mirror=False, mean=IMG_MEAN,
set='val'),
batch_size=args.batch_size, shuffle=False, num_workers=args.num_workers,
pin_memory=True)
elif args.data_ID == 8:
# real2real_City2Mapillary2IDD
seg_restore_from = './pretrained_model/MT-KD/real2real_City2Mapillary/Seg_batch_12000_miou_7527.pth'
D1_restore_from = './pretrained_model/MT-KD/real2real_City2Mapillary/D_net1_batch_12000.pth'
D2_restore_from = './pretrained_model/MT-KD/real2real_City2Mapillary/D_net2_batch_12000.pth'
data_dir_target = '/iarai/home/yonghao.xu/Data/SegmentationData/IDD/IDD_Segmentation/'
data_list_val = './dataset/IDD_val.txt'
data_list_target = './dataset/IDD_train.txt'
input_size_test = (1920,1080)
snapshot_dir = args.snapshot_dir+'real2real_City2Mapillary2IDD/UTKD_out_weight_'+str(args.out_weight)+'_con_weight_'+str(args.con_weight)+'_time'+time.strftime('%m%d_%H%M', time.localtime(time.time()))+'/'
tgt_loader = data.DataLoader(
IDDDataSet(data_dir_target, data_list_target, max_iters=args.num_steps_stop*args.batch_size,
crop_size=input_size_tgt,
scale=False, mirror=False, mean=IMG_MEAN,
set='train'),
batch_size=args.batch_size, shuffle=True, num_workers=args.num_workers,
pin_memory=True)
val_loader = data.DataLoader(
IDDDataSet(data_dir_target, data_list_val, max_iters=None,
crop_size=input_size_tgt,
scale=False, mirror=False, mean=IMG_MEAN,
set='val'),
batch_size=args.batch_size, shuffle=False, num_workers=args.num_workers,
pin_memory=True)
elif args.data_ID == 9:
# real2real_City2IDD2Mapillary
seg_restore_from = './pretrained_model/MT-KD/real2real_City2IDD/Seg_batch_33500_miou_7245.pth'
D1_restore_from = './pretrained_model/MT-KD/real2real_City2IDD/D_net1_batch_33500.pth'
D2_restore_from = './pretrained_model/MT-KD/real2real_City2IDD/D_net2_batch_33500.pth'
data_dir_target = '/iarai/home/yonghao.xu/Data/SegmentationData/MapillaryVistas/'
data_list_val = './dataset/Mapillary_val.txt'
data_list_target = './dataset/Mapillary_train.txt'
input_size_test = (1920,1080)
snapshot_dir = args.snapshot_dir+'real2real_City2IDD2Mapillary/UTKD_out_weight_'+str(args.out_weight)+'_con_weight_'+str(args.con_weight)+'_time'+time.strftime('%m%d_%H%M', time.localtime(time.time()))+'/'
tgt_loader = data.DataLoader(
MapDataSet(data_dir_target, data_list_target, max_iters=args.num_steps_stop*args.batch_size,
crop_size=input_size_tgt,
scale=False, mirror=False, mean=IMG_MEAN,
set='train'),
batch_size=args.batch_size, shuffle=True, num_workers=args.num_workers,
pin_memory=True)
val_loader = data.DataLoader(
MapDataSet(data_dir_target, data_list_val, max_iters=None,
crop_size=input_size_tgt,
scale=False, mirror=False, mean=IMG_MEAN,
set='val'),
batch_size=args.batch_size, shuffle=False, num_workers=args.num_workers,
pin_memory=True)
if os.path.exists(snapshot_dir)==False:
os.makedirs(snapshot_dir)
f = open(snapshot_dir+'log.txt', 'w')
# Create network
student_net = DeeplabMulti(num_classes=args.num_classes)
teacher_net = DeeplabMulti(num_classes=args.num_classes)
D_net1 = Discriminator(input_channel=args.num_classes)
D_net2 = Discriminator(input_channel=args.num_classes)
saved_state_dict = torch.load(D1_restore_from)
D_net1.load_state_dict(saved_state_dict)
saved_state_dict = torch.load(D2_restore_from)
D_net2.load_state_dict(saved_state_dict)
D_net1.eval()
D_net2.eval()
for name, param in D_net1.named_parameters():
param.requires_grad=False
for name, param in D_net2.named_parameters():
param.requires_grad=False
saved_state_dict = torch.load(seg_restore_from) # pre-trained MT-KD model
teacher_net.load_state_dict(saved_state_dict)
saved_state_dict = torch.load(args.restore_from) # from scratch
student_net.load_state_dict(saved_state_dict)
for name, param in teacher_net.named_parameters():
param.requires_grad=False
student_net = student_net.cuda()
teacher_net = teacher_net.cuda()
D_net1 = D_net1.cuda()
D_net2 = D_net2.cuda()
optimizer = optim.SGD(student_net.optim_parameters(args),
lr=args.learning_rate, momentum=args.momentum, weight_decay=args.weight_decay)
optimizer.zero_grad()
student_params = list(student_net.parameters())
teacher_params = list(teacher_net.parameters())
teacher_optimizer = WeightEMA(
teacher_params,
student_params,
alpha=args.teacher_alpha,
)
interp_tgt = nn.Upsample(size=(input_size_tgt[1], input_size_tgt[0]), mode='bilinear', align_corners=True)
n_class = args.num_classes
loss_hist = np.zeros((args.num_steps_stop,7))
mIoU_hist = 20
con_loss = torch.nn.MSELoss()
bce_loss = torch.nn.BCEWithLogitsLoss()
# labels for adversarial training
source_label = 0
target_label = 1
cudnn.enabled = True
cudnn.benchmark = True
student_net.eval()
teacher_net.eval()
half_batch = args.batch_size // 2
mask_generator = mask_gen.BoxMaskGenerator(prop_range=0.5, n_boxes=1,
random_aspect_ratio=not False,
prop_by_area=not False, within_bounds=not False,
invert=not False)
for batch_index, tgt_data in enumerate(tgt_loader):
if batch_index>=args.num_steps_stop:
break
decay_adv = (1 - batch_index/args.num_steps)
tem_time = time.time()
optimizer.zero_grad()
adjust_learning_rate(optimizer,args.learning_rate,batch_index,args.num_steps)
# L_con
images, _,_, _ = tgt_data
images = images.cuda()
images_1 = images[:half_batch,:,:,:]
images_2 = images[half_batch:,:,:,:]
mask_size = images.shape[2:]
# (N,1,H,W)
batch_mix_masks = torch.from_numpy(mask_generator.generate_params(len(images_1), mask_size)).cuda().float()
# Mix images with masks
mixed_s = images_1 * (1 - batch_mix_masks) + images_2 * batch_mix_masks
_,tgt_s_output_mixed_s = student_net(mixed_s)
_,tgt_t_output_ori1 = teacher_net(images_1)
_,tgt_t_output_ori2 = teacher_net(images_2)
tgt_t_output = interp_tgt(tgt_t_output_ori1) * (1 - batch_mix_masks) + interp_tgt(tgt_t_output_ori2) * batch_mix_masks
tgt_s_output = interp_tgt(tgt_s_output_mixed_s)
tgt_t_predicts = F.softmax(tgt_t_output, dim=1).transpose(1, 2).transpose(2, 3)
tgt_s_predicts = F.softmax(tgt_s_output, dim=1).transpose(1, 2).transpose(2, 3)
tgt_s_predicts = tgt_s_predicts.contiguous().view(-1,n_class)
tgt_t_predicts = tgt_t_predicts.contiguous().view(-1,n_class)
con_loss_value = con_loss(tgt_s_predicts, tgt_t_predicts)
tgt_s_output_ori1,tgt_s_output_ori2 = student_net(images)
tgt_s_output1 = interp_tgt(tgt_s_output_ori1)
tgt_s_output2 = interp_tgt(tgt_s_output_ori2)
D_out1 = D_net1(F.softmax(tgt_s_output1, dim=1))
D_out2 = D_net2(F.softmax(tgt_s_output2, dim=1))
loss_adv_tgt1 = bce_loss(D_out1,
Variable(torch.FloatTensor(D_out1.data.size()).fill_(source_label)).cuda())
loss_adv_tgt2 = bce_loss(D_out2,
Variable(torch.FloatTensor(D_out2.data.size()).fill_(source_label)).cuda())
loss_adv_G = (loss_adv_tgt1*0.1 + loss_adv_tgt2)
# Weighted sum
total_loss = args.con_weight * con_loss_value + args.out_weight * loss_adv_G * decay_adv
loss_hist[batch_index,0] = total_loss.item()
loss_hist[batch_index,2] = con_loss_value.item()
loss_hist[batch_index,3] = loss_adv_G.item()
total_loss.backward()
optimizer.step()
teacher_optimizer.step()
batch_time = time.time()-tem_time
if (batch_index+1) % 10 == 0:
print('Iter %d/%d time: %.2f con_loss = %.3f g_loss = %.3f \n'\
%(batch_index+1,args.num_steps,batch_time,np.mean(loss_hist[batch_index-9:batch_index+1,2]),\
np.mean(loss_hist[batch_index-9:batch_index+1,3])))
f.write('Iter %d/%d time: %.2f con_loss = %.3f g_loss = %.3f \n'\
%(batch_index+1,args.num_steps,batch_time,np.mean(loss_hist[batch_index-9:batch_index+1,2]),\
np.mean(loss_hist[batch_index-9:batch_index+1,3])))
f.flush()
if (batch_index+1) % 500 == 0:
mIoU_new = test_mIoU7(f,teacher_net, val_loader, batch_index+1,input_size_test,print_per_batches=10)
# Saving the models
if mIoU_new > mIoU_hist:
f.write('Save Model\n')
print('Save Model')
model_name = 'batch_'+repr(batch_index+1)+'_miou_'+repr(int(mIoU_new*100))+'.pth'
torch.save(teacher_net.state_dict(), os.path.join(
snapshot_dir, model_name))
mIoU_hist = mIoU_new
f.close()
np.savez(snapshot_dir+'loss.npz',loss_hist=loss_hist)
if __name__ == '__main__':
main()