-
Notifications
You must be signed in to change notification settings - Fork 4
/
main_TD.py
320 lines (243 loc) · 13.6 KB
/
main_TD.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
'''
main Top-Down pruning
'''
import os
import random
import pickle
import argparse
import numpy as np
import torch
import torch.optim
import torch.nn as nn
import torch.utils.data
import torch.nn.functional as F
from utils import *
from trainer import *
from dataloader import *
from model import PreActResNet18 as ResNet18
parser = argparse.ArgumentParser(description='PyTorch CIL Top-Down pruning')
#################### base setting #########################
parser.add_argument('--data', help='The directory for data', default='data/cifar10', type=str)
parser.add_argument('--dataset', type=str, default='cifar10', help='default dataset')
parser.add_argument('--save_dir', help='The directory used to save the trained models', default='TD_cifar10', type=str)
parser.add_argument('--save_data_path', help='The directory used to save the data', default='TD_cifar10/data', type=str)
parser.add_argument('--print_freq', default=50, type=int, help='print frequency')
parser.add_argument('--gpu', type=int, default=0, help='gpu device id')
parser.add_argument('--seed', type=int, default=None, help='random seed')
################## training setting ###########################
parser.add_argument('--epochs', default=100, type=int, help='number of total epochs to run')
parser.add_argument('--batch_size', default=256, type=int, help='batch size')
parser.add_argument('--lr', default=0.01, type=float, help='initial learning rate')
parser.add_argument('--decreasing_lr', default='60,80', help='decreasing strategy')
parser.add_argument('--momentum', default=0.9, type=float, help='momentum')
parser.add_argument('--weight_decay', default=5e-4, type=float, help='weight decay')
################## CIL setting ##################################
parser.add_argument('--classes_per_classifier', type=int, default=2, help='number of classes per classifier')
parser.add_argument('--classifiers', type=int, default=5, help='number of classifiers')
parser.add_argument('--unlabel_num', type=int, default=50, help='number of unlabel images')
################## pruning setting ##################################
parser.add_argument('--iter_epochs', default=30, type=int, help='number retrain-epoch during iterative pruning')
parser.add_argument('--percent', default=0.2, type=float, help='pruning rate')
parser.add_argument('--rewind', type=str, default='zero', help='rewind_type')
parser.add_argument('--prune_scheduler', default='1,1,2,3,5', help='pruning times for each task')
best_prec1 = 0
def main():
global args, best_prec1
args = parser.parse_args()
print(args)
#pre-define pruning schedule
prune_steps = [x-1 for x in list(map(int, args.prune_scheduler.split(',')))]
assert len(prune_steps) == args.classifiers
pruning_flag = False
prune_stage = 0
decreasing_lr = list(map(int, args.decreasing_lr.split(',')))
all_states = args.classifiers
class_per_state = args.classes_per_classifier
torch.cuda.set_device(int(args.gpu))
if args.seed:
setup_seed(args.seed)
os.makedirs(args.save_dir, exist_ok=True)
os.makedirs(args.save_data_path, exist_ok=True)
#setup logger
log_result = Logger(os.path.join(args.save_dir, 'log_results.txt'))
name_list = ['Task{}'.format(i+1) for i in range(all_states)]
name_list.append('Mean Acc')
log_result.append(['current state = 1'])
log_result.append(name_list)
criterion = nn.CrossEntropyLoss()
model = ResNet18(num_classes_per_classifier=class_per_state, num_classifier=all_states)
model.cuda()
torch.save({
'state_dict': model.state_dict(),
}, os.path.join(args.save_dir, 'task0_checkpoint_weight.pt'))
train_loader, val_loader = setup_dataset(args, task_id=0, train=True)
optimizer = torch.optim.SGD(model.parameters(), args.lr,
momentum=args.momentum,
weight_decay=args.weight_decay)
scheduler = torch.optim.lr_scheduler.MultiStepLR(optimizer, milestones=decreasing_lr, gamma=0.1)
for epoch in range(args.epochs):
print("The learning rate is {}".format(optimizer.param_groups[0]['lr']))
train_accuracy = train(train_loader, model, criterion, optimizer, epoch, args)
prec1 = validate(val_loader, model, criterion, args, fc_num=1, if_main=True)
scheduler.step()
is_best = prec1 > best_prec1
best_prec1 = max(prec1, best_prec1)
save_checkpoint({
'epoch': epoch + 1,
'state_dict': model.state_dict(),
'best_prec1': best_prec1,
'optimizer': optimizer,
}, is_best, args.save_dir, filename='task1_checkpoint.pt', best_name='task1_best_model.pt')
for current_state in range(1, all_states+1):
best_prec1 = 0
model = ResNet18(num_classes_per_classifier=class_per_state, num_classifier=all_states)
model.cuda()
model_path = os.path.join(args.save_dir, 'task'+str(current_state)+'_best_model.pt')
new_dict = torch.load(model_path, map_location=torch.device('cuda:'+str(args.gpu)))
if pruning_flag:
print('pruning with custom mask')
current_mask = extract_mask(new_dict['state_dict'])
prune_model_custom(model, current_mask)
remain_weight = check_sparsity(model)
model.load_state_dict(new_dict['state_dict'])
print('*****************************************************************************')
print('start training task'+str(current_state+1))
print('best epoch', new_dict['epoch'])
print('model loaded', model_path)
print('remain weight size = {}'.format(remain_weight))
print('*****************************************************************************')
bal_acc = []
log_acc = ['None' for i in range(all_states+1)]
for test_iter in range(current_state):
test_loader = setup_dataset(args, task_id=test_iter, train=False)
ta_bal = validate(test_loader, model, criterion, args, fc_num = current_state, if_main= True)
bal_acc.append(ta_bal)
log_acc[test_iter] = ta_bal
print('* test accuracy for data {0} = {1:.2f} '.format(test_iter+1, ta_bal))
mean_acc = np.mean(np.array(bal_acc))
log_acc[-1] = mean_acc
print('******************************************************')
print('* mean accuracy for state {0} = {1:.2f} '.format(current_state, mean_acc))
print('******************************************************')
log_result.append(log_acc)
log_result.append(['remain weight size = {:.4f}'.format(remain_weight)])
log_result.append(['*'*50])
log_result.append(['current state = {}'.format(current_state+1)])
log_result.append(name_list)
generate_softlogit_unlabel(args, current_state, model, criterion)
#pruning stage
last_checkpoint_weight = torch.load(os.path.join(args.save_dir, 'task'+str(current_state)+'_checkpoint.pt'), map_location=torch.device('cuda:'+str(args.gpu)))
model.load_state_dict(last_checkpoint_weight['state_dict'])
if prune_steps[prune_stage] > 0:
# iterative pruning
if current_state == 1:
train_loader, _ = setup_dataset(args, task_id=0, train=True)
optimizer = torch.optim.SGD(model.parameters(), args.lr/100,
momentum=args.momentum,
weight_decay=args.weight_decay)
for prune_iter in range(prune_steps[prune_stage]):
print('starting pruning', prune_steps[prune_stage])
pruning_model(model, args.percent)
check_sparsity(model)
pruning_flag = True
for epoch in range(args.iter_epochs):
train(train_loader, model, criterion, optimizer, epoch, args)
else:
train_loader_random, train_loader_balance_new, train_loader_balance_old, unlabel_loader, _ = setup_dataset(args, current_state-1, train=True)
optimizer = torch.optim.SGD(model.parameters(), args.lr/100,
momentum=args.momentum,
weight_decay=args.weight_decay)
for prune_iter in range(prune_steps[prune_stage]):
print('starting pruning', prune_steps[prune_stage])
pruning_model(model,args.percent)
check_sparsity(model)
pruning_flag = True
for epoch in range(args.iter_epochs):
train_KD(train_loader_random, train_loader_balance_new, train_loader_balance_old, unlabel_loader, model, criterion, optimizer, epoch, current_state, args)
if prune_steps[prune_stage] >= 0:
pruning_model(model, args.percent)
check_sparsity(model)
pruning_flag = True
prune_stage += 1
#rewind
if args.rewind == 'best':
print('rewind best weight')
model_path = os.path.join(args.save_dir, 'task'+str(current_state)+'_best_model.pt')
new_dict = torch.load(model_path, map_location=torch.device('cuda:'+str(args.gpu)))['state_dict']
weight_orig_dict = rewind(model, new_dict, pruning_flag)
model.load_state_dict(weight_orig_dict, strict=False)
elif args.rewind == 'zero':
print('rewind zero weight')
model_path = os.path.join(args.save_dir, 'task0_checkpoint_weight.pt')
new_dict = torch.load(model_path, map_location=torch.device('cuda:'+str(args.gpu)))['state_dict']
weight_orig_dict = rewind(model, new_dict, pruning_flag)
model.load_state_dict(weight_orig_dict, strict=False)
elif args.rewind == 'rand':
print('random re-init')
new_dict = ResNet18(num_classes_per_classifier=class_per_state, num_classifier=all_states).state_dict()
weight_orig_dict = rewind(model, new_dict, pruning_flag)
model.load_state_dict(weight_orig_dict, strict=False)
else:
print('finetune')
if current_state == all_states:
print('re-train task{}'.format(current_state))
current_state -= 1
save_state = current_state+2
else:
save_state = current_state+1
train_loader_random, train_loader_balance_new, train_loader_balance_old, unlabel_loader, val_loader = setup_dataset(args, current_state, train=True)
optimizer = torch.optim.SGD(model.parameters(), args.lr,
momentum=args.momentum,
weight_decay=args.weight_decay)
scheduler = torch.optim.lr_scheduler.MultiStepLR(optimizer, milestones=decreasing_lr, gamma=0.1)
for epoch in range(args.epochs):
print("The learning rate is {}".format(optimizer.param_groups[0]['lr']))
train_accuracy = train_KD(train_loader_random, train_loader_balance_new, train_loader_balance_old, unlabel_loader, model, criterion, optimizer, epoch, current_state+1, args)
prec1 = validate(val_loader, model, criterion, args, fc_num=current_state+1, if_main=True)
scheduler.step()
# remember best prec@1 and save checkpoint
is_best = prec1 > best_prec1
best_prec1 = max(prec1, best_prec1)
save_checkpoint({
'epoch': epoch + 1,
'state_dict': model.state_dict(),
'best_prec1': best_prec1,
'optimizer': optimizer,
}, is_best, args.save_dir, filename='task{}_checkpoint.pt'.format(save_state), best_name='task{}_best_model.pt'.format(save_state))
if current_state == all_states:
current_state += 1
# test
current_state = all_states+1
model = ResNet18(num_classes_per_classifier=class_per_state, num_classifier=all_states)
model.cuda()
model_path = os.path.join(args.save_dir, 'task'+str(current_state)+'_best_model.pt')
new_dict = torch.load(model_path, map_location=torch.device('cuda:'+str(args.gpu)))
if pruning_flag:
print('pruning with custom mask')
current_mask = extract_mask(new_dict['state_dict'])
prune_model_custom(model, current_mask)
model.load_state_dict(new_dict['state_dict'])
remain_weight = check_sparsity(model)
print('*****************************************************************************')
print('start testing task'+str(current_state))
print('remain weight size = {}'.format(remain_weight))
print('*****************************************************************************')
# testing accuracy & generate feature of unlabeled data using original model
bal_acc = []
log_acc = ['None' for i in range(all_states+1)]
for test_iter in range(all_states):
test_loader = setup_dataset(args, task_id=test_iter, train=False)
ta_bal = validate(test_loader, model, criterion, args, fc_num = all_states, if_main= True)
bal_acc.append(ta_bal)
log_acc[test_iter] = ta_bal
print('* test accuracy for data {0} = {1:.2f} '.format(test_iter+1, ta_bal))
mean_acc = np.mean(np.array(bal_acc))
log_acc[-1] = mean_acc
print('******************************************************')
print('* mean accuracy for state {0} = {1:.2f} '.format(all_states, mean_acc))
print('******************************************************')
log_result.append(log_acc)
log_result.append(['remain weight size = {:.4f}'.format(remain_weight)])
log_result.append(['*'*50])
if __name__ == '__main__':
main()