-
Notifications
You must be signed in to change notification settings - Fork 0
/
main.py
269 lines (232 loc) · 13.5 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
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
import os
import sys
import argparse
import torch
import numpy as np
from random import shuffle
from collections import OrderedDict
from dataloaders.datasetGen import SplitGen, PermutedGen
from utils.utils import factory
import random
import pdb
def run(args):
if not os.path.exists('outputs'):
os.mkdir('outputs')
# Prepare dataloaders
# train_dataset, val_dataset = dataloaders.base.__dict__[args.dataset](args.dataroot, args.train_aug)
train_dataset, val_dataset = factory(
'dataloaders', 'base', args.dataset)(args.dataroot, args.train_aug)
if args.n_permutation > 0:
train_dataset_splits, val_dataset_splits, task_output_space = PermutedGen(train_dataset, val_dataset,
args.n_permutation,
remap_class=not args.no_class_remap)
else:
train_dataset_splits, val_dataset_splits, task_output_space = SplitGen(train_dataset, val_dataset,
first_split_sz=args.first_split_size,
other_split_sz=args.other_split_size,
rand_split=args.rand_split,
remap_class=not args.no_class_remap)
# Prepare the Agent (model)
dataset_name = args.dataset + \
'_{}'.format(args.first_split_size) + \
'_{}'.format(args.other_split_size)
agent_config = {'model_lr': args.model_lr, 'momentum': args.momentum, 'model_weight_decay': args.model_weight_decay,
'schedule': args.schedule,
'schedule_pruned': args.schedule_pruned,
'model_type': args.model_type, 'model_name': args.model_name, 'model_weights': args.model_weights,
'out_dim': {'All': args.force_out_dim} if args.force_out_dim > 0 else task_output_space,
'model_optimizer': args.model_optimizer,
'print_freq': args.print_freq,
'gpu': True if args.gpuid[0] >= 0 else False,
'with_head': args.with_head,
'reset_model_opt': args.reset_model_opt,
'reg_coef': args.reg_coef,
'head_lr': args.head_lr,
'svd_lr': args.svd_lr,
'bn_lr': args.bn_lr,
'svd_thres': args.svd_thres,
'gamma': args.gamma,
'dataset_name': dataset_name,
'regularization': args.sr,
'regularization_weight': args.s,
'pruning': args.p,
'pruning_ratio': args.prun_ratio
}
# agent = agents.__dict__[args.agent_type].__dict__[args.agent_name](agent_config)
agent = factory('svd_agent', args.agent_type,
args.agent_name)(agent_config)
# Decide split ordering
task_names = sorted(list(task_output_space.keys()), key=int)
print('Task order:', task_names)
acc_table = OrderedDict()
acc_table_train = OrderedDict()
for i in range(len(task_names)):
train_name = task_names[i]
print('======================', train_name,
'=======================')
train_loader = torch.utils.data.DataLoader(train_dataset_splits[train_name],
batch_size=args.batch_size, shuffle=True,
num_workers=args.workers)
val_loader = torch.utils.data.DataLoader(val_dataset_splits[train_name],
batch_size=args.batch_size, shuffle=False,
num_workers=args.workers)
if args.incremental_class:
agent.add_valid_output_dim(task_output_space[train_name])
# Learn
agent.train_task(train_loader, val_loader, train_name)
torch.cuda.empty_cache()
# Evaluate
acc_table[train_name] = OrderedDict()
acc_table_train[train_name] = OrderedDict()
for j in range(i + 1):
val_name = task_names[j]
# store the current model's bn's parameters and temporality replace them with the stored ones
original_bn_parameters = agent.change_bn_parameters(val_name)
print('validation split name:', val_name)
val_data = val_dataset_splits[val_name] if not args.eval_on_train_set else train_dataset_splits[
val_name]
val_loader = torch.utils.data.DataLoader(val_data,
batch_size=args.batch_size, shuffle=False,
num_workers=args.workers)
acc_table[val_name][train_name] = agent.validation(val_loader)
print("**************************************************")
print('training split name:', val_name)
train_data = train_dataset_splits[val_name] if not args.eval_on_train_set else train_dataset_splits[
val_name]
train_loader = torch.utils.data.DataLoader(train_data,
batch_size=args.batch_size, shuffle=False,
num_workers=args.workers)
acc_table_train[val_name][train_name] = agent.validation(
train_loader)
print("**************************************************")
# restore the bn's parameters
agent.recover_bn_parameters(original_bn_parameters)
return acc_table, task_names
def get_args(argv):
# This function prepares the variables shared across demo.py
parser = argparse.ArgumentParser()
parser.add_argument('--gpuid', nargs="+", type=int, default=[1],
help="The list of gpuid, ex:--gpuid 3 1. Negative value means cpu-only")
parser.add_argument('--model_type', type=str, default='resnet',
help="The type (mlp|lenet|vgg|resnet) of backbone network")
parser.add_argument('--model_name', type=str, default='resnet18',
help="The name of actual model for the backbone")
parser.add_argument('--force_out_dim', type=int, default=0,
help="Set 0 to let the task decide the required output dimension")
parser.add_argument('--agent_type', type=str,
default='svd_based', help="The type (filename) of agent")
parser.add_argument('--agent_name', type=str,
default='svd_based', help="The class name of agent")
parser.add_argument('--model_optimizer', type=str, default='Adam',
help="SGD|Adam|RMSprop|amsgrad|Adadelta|Adagrad|Adamax ...")
parser.add_argument('--dataroot', type=str, default='../data',
help="The root folder of dataset or downloaded data")
parser.add_argument('--dataset', type=str, default='CIFAR100',
help="MNIST(default)|CIFAR10|CIFAR100")
parser.add_argument('--n_permutation', type=int, default=0,
help="Enable permuted tests when >0")
parser.add_argument('--first_split_size', type=int, default=10)
parser.add_argument('--other_split_size', type=int, default=10)
parser.add_argument('--no_class_remap', dest='no_class_remap', default=False, action='store_true',
help="Avoid the dataset with a subset of classes doing the remapping. Ex: [2,5,6 ...] -> [0,1,2 ...]") # class:we need to know specific class,other:no need to know specific class
parser.add_argument('--train_aug', dest='train_aug', default=True, action='store_false',
help="Allow data augmentation during training")
parser.add_argument('--rand_split', dest='rand_split', default=False, action='store_true',
help="Randomize the classes in splits")
parser.add_argument('--workers', type=int, default=0,
help="#Thread for dataloader")
parser.add_argument('--batch_size', type=int, default=128)
parser.add_argument('--model_lr', type=float,
default=0.0005, help="Classifier Learning rate")
parser.add_argument('--head_lr', type=float,
default=0.0005, help="Classifier Learning rate")
parser.add_argument('--svd_lr', type=float, default=0.0005,
help="Classifier Learning rate")
parser.add_argument('--bn_lr', type=float, default=0.0005,
help="Classifier Learning rate")
parser.add_argument('--gamma', type=float, default=0.5,
help="Learning rate decay")
parser.add_argument('--svd_thres', type=float,
default=1.0, help='reserve eigenvector')
parser.add_argument('--momentum', type=float, default=0)
parser.add_argument('--model_weight_decay',
type=float, default=1e-5) # 1e-4
parser.add_argument('--schedule', nargs="+", type=int, default=[1],
help="epoch ")
parser.add_argument('--schedule_pruned', nargs="+", type=int, default=[1],
help="epoch for retraining pruned model")
parser.add_argument('--print_freq', type=float, default=10,
help="Print the log at every x iteration")
parser.add_argument('--model_weights', type=str, default=None,
help="The path to the file for the model weights (*.pth).")
parser.add_argument('--eval_on_train_set', dest='eval_on_train_set', default=False, action='store_true',
help="Force the evaluation on train set")
parser.add_argument('--offline_training', dest='offline_training', default=False, action='store_true',
help="Non-incremental learning by make all data available in one batch. For measuring the upperbound performance.")
parser.add_argument('--repeat', type=int, default=1,
help="Repeat the experiment N times")
parser.add_argument('--incremental_class', dest='incremental_class', default=False, action='store_true',
help="The number of output node in the single-headed model increases along with new categories.")
parser.add_argument('--with_head', dest='with_head', default=False, action='store_true',
help="whether constraining head")
parser.add_argument('--reset_model_opt', dest='reset_model_opt', default=True, action='store_true',
help="whether reset optimizer for model at the start of training each tasks")
parser.add_argument('--reg_coef', type=float, default=100,
help="The coefficient for ewc reg")
parser.add_argument('--sparsity-regularization', '-sr', dest='sr', action='store_true',
help='train with channel sparsity regularization')
parser.add_argument('--prun', '-p', dest='p', action='store_true',
help='learning in the pruned null space')
parser.add_argument('--s', type=float, default=0.0001,
help='scale sparse rate (default: 0.0001)')
parser.add_argument('--prun_ratio', type=int, default=30,
help='percentage of pruned parameters (default: 30)')
args = parser.parse_args(argv)
return args
if __name__ == '__main__':
args = get_args(sys.argv[1:])
avg_final_acc = np.zeros(args.repeat)
final_bwt = np.zeros(args.repeat)
torch.cuda.set_device(args.gpuid[0])
# Seed
SEED = 0
np.random.seed(SEED)
torch.manual_seed(SEED)
random.seed(SEED)
torch.cuda.manual_seed(SEED)
torch.backends.cudnn.deterministic = True
torch.backends.cudnn.benchmark = False
if torch.cuda.is_available():
torch.cuda.manual_seed(SEED)
for r in range(args.repeat):
# Run the experiment
acc_table, task_names = run(args)
print(acc_table)
# Calculate average performance across tasks
# Customize this part for a different performance metric
avg_acc_history = [0] * len(task_names)
bwt_history = [0] * len(task_names)
for i in range(len(task_names)):
train_name = task_names[i]
cls_acc_sum = 0
backward_transfer = 0
for j in range(i + 1):
val_name = task_names[j]
cls_acc_sum += acc_table[val_name][train_name]
backward_transfer += acc_table[val_name][train_name] - \
acc_table[val_name][val_name]
avg_acc_history[i] = cls_acc_sum / (i + 1)
bwt_history[i] = backward_transfer / i if i > 0 else 0
print('Task', train_name, 'average acc:', avg_acc_history[i])
print('Task', train_name, 'backward transfer:', bwt_history[i])
# Gather the final avg accuracy
avg_final_acc[r] = avg_acc_history[-1]
final_bwt[r] = bwt_history[-1]
# Print the summary so far
print('===Summary of experiment repeats:',
r + 1, '/', args.repeat, '===')
print('The last avg acc of all repeats:', avg_final_acc)
print('The last bwt of all repeats:', final_bwt)
print('acc mean:', avg_final_acc.mean(),
'acc std:', avg_final_acc.std())
print('bwt mean:', final_bwt.mean(), 'bwt std:', final_bwt.std())