-
Notifications
You must be signed in to change notification settings - Fork 1
/
opmatch_trainer.py
378 lines (316 loc) · 15.2 KB
/
opmatch_trainer.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
import logging
import time
import copy
import os
import numpy as np
import torch
import shutil
import torch.nn.functional as F
from torch.utils.data import DataLoader, RandomSampler, Subset
from torch.utils.data.distributed import DistributedSampler
from tqdm import tqdm
from utils.opmatch_misc import AverageMeter, ova_loss, ova_ent, exclude_dataset, accuracy
from torchvision import transforms
from utils.randaugment import RandAugmentMC
logger = logging.getLogger(__name__)
best_acc = 0
best_acc_val = 0
def save_checkpoint(state, is_best, checkpoint, filename='checkpoint.pth.tar'):
filepath = os.path.join(checkpoint, filename)
torch.save(state, filepath)
if is_best:
shutil.copyfile(filepath, os.path.join(checkpoint, 'model_best.pth.tar'))
class TransformFixMatch(object):
def __init__(self, mean, std, norm=True, size_image=32):
self.weak = transforms.Compose([
transforms.Resize(size_image),
transforms.RandomHorizontalFlip(),
transforms.RandomCrop(size=size_image,
padding=int(size_image * 0.125),
padding_mode='reflect')])
self.weak2 = transforms.Compose([
transforms.Resize(size_image),
transforms.RandomHorizontalFlip(), ])
self.strong = transforms.Compose([
transforms.Resize(size_image),
transforms.RandomHorizontalFlip(),
transforms.RandomCrop(size=size_image,
padding=int(size_image * 0.125),
padding_mode='reflect'),
RandAugmentMC(n=2, m=10)])
self.normalize = transforms.Compose([
transforms.ToTensor(),
transforms.Normalize(mean=mean, std=std)])
self.norm = norm
def __call__(self, x):
weak = self.weak(x)
strong = self.strong(x)
if self.norm:
return self.normalize(weak), self.normalize(strong), self.normalize(self.weak2(x))
else:
return weak, strong
class TransformOpenMatch(object):
def __init__(self, mean, std, norm=True, size_image=32):
self.weak = transforms.Compose([
transforms.Resize(size_image),
transforms.RandomHorizontalFlip(),
transforms.RandomCrop(size=size_image,
padding=int(size_image * 0.125),
padding_mode='reflect')])
self.weak2 = transforms.Compose([
transforms.Resize(size_image),
transforms.RandomHorizontalFlip(), ])
self.normalize = transforms.Compose([
transforms.ToTensor(),
transforms.Normalize(mean=mean, std=std)])
self.norm = norm
def __call__(self, x):
weak = self.weak(x)
strong = self.weak(x)
if self.norm:
return self.normalize(weak), self.normalize(strong), self.normalize(self.weak2(x))
else:
return weak, strong
def train(args, labeled_dataset, unlabeled_dataset, test_loader, model, optimizer, ema_model, scheduler):
print('**************** labeled data %d unlabeled data %d' % (len(labeled_dataset), len(unlabeled_dataset)))
if args.amp:
from apex import amp
global best_acc
test_accs = []
batch_time = AverageMeter()
data_time = AverageMeter()
losses = AverageMeter()
losses_x = AverageMeter()
losses_o = AverageMeter()
losses_oem = AverageMeter()
losses_socr = AverageMeter()
losses_fix = AverageMeter()
mask_probs = AverageMeter()
end = time.time()
default_out = "Epoch: {epoch}/{epochs:4}. " \
"LR: {lr:.6f}. " \
"Lab: {loss_x:.4f}. " \
"Open: {loss_o:.4f}"
output_args = vars(args)
default_out += " OEM {loss_oem:.4f}"
default_out += " SOCR {loss_socr:.4f}"
default_out += " Fix {loss_fix:.4f}"
cifar100_mean = (0.5071, 0.4867, 0.4408)
cifar100_std = (0.2675, 0.2565, 0.2761)
train_sampler = RandomSampler if args.local_rank == -1 else DistributedSampler
# unlabeled data for ood detection
unlabeled_dataset.transform = TransformFixMatch(mean=cifar100_mean, std=cifar100_std)
unlabeled_trainloader = DataLoader(unlabeled_dataset, sampler=train_sampler(unlabeled_dataset),
batch_size=args.batch_size * args.mu, num_workers=args.num_workers,
drop_last=True)
unlabeled_iter = iter(unlabeled_trainloader)
# all unlabeled data for auxilary
unlabeled_dataset_all = copy.deepcopy(unlabeled_dataset)
unlabeled_dataset_all.transform = TransformOpenMatch(mean=cifar100_mean, std=cifar100_std)
unlabeled_trainloader_all = DataLoader(unlabeled_dataset_all, sampler=train_sampler(unlabeled_dataset_all),
batch_size=args.batch_size * args.mu, num_workers=args.num_workers,
drop_last=True)
unlabeled_all_iter = iter(unlabeled_trainloader_all)
# labeled data
labeled_dataset.transform = TransformOpenMatch(mean=cifar100_mean, std=cifar100_std)
labeled_trainloader = DataLoader(labeled_dataset, sampler=train_sampler(labeled_dataset),
batch_size=args.batch_size, num_workers=args.num_workers, drop_last=True)
labeled_iter = iter(labeled_trainloader)
if args.world_size > 1:
labeled_epoch = 0
unlabeled_epoch = 0
use_fix = True
for epoch in range(args.start_epoch, args.epochs):
model.train()
output_args["epoch"] = epoch
if not args.no_progress:
p_bar = tqdm(range(args.eval_step))
if epoch >= args.start_fix:
# TODO check
sel_inndex = exclude_dataset(args, unlabeled_dataset, ema_model.ema)
if len(sel_inndex) <= args.batch_size * args.mu:
# unlabeled_trainloader = DataLoader(unlabeled_dataset, sampler=train_sampler(unlabeled_dataset),
# batch_size=len(unlabeled_dataset), num_workers=args.num_workers,
# drop_last=True)
use_fix = False
# assert NotImplementedError
else:
unlabeled_dataset = Subset(unlabeled_dataset, sel_inndex)
unlabeled_trainloader = DataLoader(unlabeled_dataset, sampler=train_sampler(unlabeled_dataset),
batch_size=args.batch_size * args.mu, num_workers=args.num_workers,
drop_last=True)
unlabeled_iter = iter(unlabeled_trainloader)
for batch_idx in range(args.eval_step):
## labeled data
try:
(_, inputs_x_s, inputs_x), targets_x, _ = labeled_iter.next()
except:
if args.world_size > 1:
labeled_epoch += 1
labeled_trainloader.sampler.set_epoch(labeled_epoch)
labeled_iter = iter(labeled_trainloader)
(_, inputs_x_s, inputs_x), targets_x, _ = labeled_iter.next()
## unlabeled data
if use_fix:
try:
(inputs_u_w, inputs_u_s, _), _ = unlabeled_iter.next()
except:
if args.world_size > 1:
unlabeled_epoch += 1
unlabeled_trainloader.sampler.set_epoch(unlabeled_epoch)
unlabeled_iter = iter(unlabeled_trainloader)
(inputs_u_w, inputs_u_s, _), _ = unlabeled_iter.next()
## all unlabeled data
try:
(inputs_all_w, inputs_all_s, _), _ = unlabeled_all_iter.next()
except:
unlabeled_all_iter = iter(unlabeled_trainloader_all)
(inputs_all_w, inputs_all_s, _), _ = unlabeled_all_iter.next()
data_time.update(time.time() - end)
b_size = inputs_x.shape[0]
# binary classification on all data
inputs_all = torch.cat([inputs_all_w, inputs_all_s], 0)
inputs = torch.cat([inputs_x, inputs_x_s, inputs_all], 0).to(args.device)
targets_x = targets_x.to(args.device)
## Feed data
logits, logits_open = model(inputs, is_feat=True)
logits_open_u1, logits_open_u2 = logits_open[2 * b_size:].chunk(2)
# for labeled data
## Loss for labeled samples
Lx = F.cross_entropy(logits[:2 * b_size], targets_x.repeat(2), reduction='mean')
# build the binary classification
Lo = ova_loss(logits_open[:2 * b_size], targets_x.repeat(2))
# for unlabeled data
## Open-set entropy minimization
L_oem = ova_ent(logits_open_u1) / 2.
L_oem += ova_ent(logits_open_u2) / 2.
## Soft consistenty regularization
logits_open_u1 = logits_open_u1.view(logits_open_u1.size(0), 2, -1)
logits_open_u2 = logits_open_u2.view(logits_open_u2.size(0), 2, -1)
logits_open_u1 = F.softmax(logits_open_u1, 1)
logits_open_u2 = F.softmax(logits_open_u2, 1)
L_socr = torch.mean(torch.sum(torch.sum(torch.abs(logits_open_u1 - logits_open_u2) ** 2, 1), 1))
if epoch >= args.start_fix and use_fix:
## pseduo label only on part data
inputs_ws = torch.cat([inputs_u_w, inputs_u_s], 0).to(args.device)
logits, _ = model(inputs_ws, is_feat=True)
logits_u_w, logits_u_s = logits.chunk(2)
## weak guide strong
pseudo_label = torch.softmax(logits_u_w.detach() / args.T, dim=-1)
max_probs, targets_u = torch.max(pseudo_label, dim=-1)
mask = max_probs.ge(args.threshold).float()
L_fix = (F.cross_entropy(logits_u_s, targets_u, reduction='none') * mask).mean()
mask_probs.update(mask.mean().item())
else:
L_fix = torch.zeros(1).to(args.device).mean()
loss = Lx + Lo + args.lambda_oem * L_oem + args.lambda_socr * L_socr + args.lambda_fix * L_fix
losses.update(loss.item())
losses_x.update(Lx.item())
losses_o.update(Lo.item())
losses_oem.update(L_oem.item())
losses_socr.update(L_socr.item())
losses_fix.update(L_fix.item())
output_args["batch"] = batch_idx
output_args["loss_x"] = losses_x.avg
output_args["loss_o"] = losses_o.avg
output_args["loss_oem"] = losses_oem.avg
output_args["loss_socr"] = losses_socr.avg
output_args["loss_fix"] = losses_fix.avg
output_args["used_un"] = len(unlabeled_dataset)
output_args["lr"] = [group["lr"] for group in optimizer.param_groups][0]
optimizer.zero_grad()
if args.amp:
# half precision
with amp.scale_loss(loss, optimizer) as scaled_loss:
scaled_loss.backward()
else:
loss.backward()
optimizer.step()
if args.opt != 'adam':
scheduler.step()
if args.use_ema:
ema_model.update(model)
batch_time.update(time.time() - end)
end = time.time()
if not args.no_progress:
p_bar.set_description(default_out.format(**output_args))
p_bar.update()
if not args.no_progress:
p_bar.close()
if args.use_ema:
test_model = ema_model.ema
else:
test_model = model
if args.local_rank in [-1, 0]:
test_loss, test_acc = test(args, test_loader, test_model, epoch)
args.writer.add_scalar('train/1.train_loss', losses.avg, epoch)
args.writer.add_scalar('train/2.train_loss_x', losses_x.avg, epoch)
args.writer.add_scalar('train/3.train_loss_o', losses_o.avg, epoch)
args.writer.add_scalar('train/4.train_loss_oem', losses_oem.avg, epoch)
args.writer.add_scalar('train/5.train_loss_socr', losses_socr.avg, epoch)
args.writer.add_scalar('train/5.train_loss_fix', losses_fix.avg, epoch)
args.writer.add_scalar('train/6.mask', mask_probs.avg, epoch)
args.writer.add_scalar('train/7.used_un', len(unlabeled_dataset), epoch)
args.writer.add_scalar('test/1.test_acc', test_acc, epoch)
args.writer.add_scalar('test/2.test_loss', test_loss, epoch)
is_best = test_acc > best_acc
best_acc = max(test_acc, best_acc)
model_to_save = model.module if hasattr(model, "module") else model
if args.use_ema:
ema_to_save = ema_model.ema.module if hasattr(
ema_model.ema, "module") else ema_model.ema
save_checkpoint({
'epoch': epoch + 1,
'model_state_dict': model_to_save.state_dict(),
'ema_state_dict': ema_to_save.state_dict() if args.use_ema else None,
'acc': test_acc,
'best_acc': best_acc, }, is_best, args.out)
test_accs.append(test_acc)
print('Best top-1 acc(test): {:.2f}'.format(best_acc))
print('Mean top-1 acc(test): {:.2f}'.format(np.mean(test_accs[-20:])))
print('curr top-1 acc(test): {:.2f}'.format(test_acc))
if args.local_rank in [-1, 0]:
args.writer.close()
with open(args.out + '/res_%s.txt'%str(time.ctime()), 'w') as f:
f.write('%.4f' % best_acc)
f.write('\n')
f.write('%.4f' % np.mean(test_accs[-20:]))
def test(args, test_loader, model, epoch):
batch_time = AverageMeter()
data_time = AverageMeter()
losses = AverageMeter()
top1 = AverageMeter()
top5 = AverageMeter()
end = time.time()
model.eval()
if not args.no_progress:
test_loader = tqdm(test_loader,
disable=args.local_rank not in [-1, 0])
with torch.no_grad():
for batch_idx, (inputs, targets) in enumerate(test_loader):
data_time.update(time.time() - end)
inputs = inputs.to(args.device)
targets = targets.to(args.device)
outputs = model(inputs)
loss = F.cross_entropy(outputs, targets)
prec1, prec5 = accuracy(outputs, targets, topk=(1, 5))
losses.update(loss.item(), inputs.shape[0])
top1.update(prec1.item(), inputs.shape[0])
top5.update(prec5.item(), inputs.shape[0])
batch_time.update(time.time() - end)
end = time.time()
if not args.no_progress:
test_loader.set_description(
"Test Iter: {batch:4}/{iter:4}. Data: {data:.3f}s. Batch: {bt:.3f}s. Loss: {loss:.4f}. top1: {top1:.2f}. top5: {top5:.2f}. ".format(
batch=batch_idx + 1,
iter=len(test_loader),
data=data_time.avg,
bt=batch_time.avg,
loss=losses.avg,
top1=top1.avg,
top5=top5.avg,
))
if not args.no_progress:
test_loader.close()
model.train()
return losses.avg, top1.avg