-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathpfge.py
166 lines (142 loc) · 7.08 KB
/
pfge.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
import argparse
import numpy as np
import os
import sys
import tabulate
import time
import torch
import torch.nn.functional as F
import data
import models
import utils
parser = argparse.ArgumentParser(description='PFGE training')
parser.add_argument('--dir', type=str, default=None, metavar='DIR',
help='training directory (default: /tmp/pfge)')
parser.add_argument('--dataset', type=str, default='CIFAR10', metavar='DATASET',
help='dataset name (default: CIFAR10)')
parser.add_argument('--use_test', action='store_true', default=True,
help='switches between validation and test set (default: validation)')
parser.add_argument('--data_path', type=str, default=None, metavar='PATH',
help='path to datasets location (default: None)')
parser.add_argument('--batch_size', type=int, default=128, metavar='N',
help='input batch size (default: 128)')
parser.add_argument('--num-workers', type=int, default=4, metavar='N',
help='number of workers (default: 4)')
parser.add_argument("--split_classes", type=int, default=None)
parser.add_argument('--model', type=str, default='VGG16', metavar='MODEL',
help='model name (default: None)')
parser.add_argument('--ckpt', type=str, default=None, metavar='CKPT',
help='checkpoint to eval (default: None)')
parser.add_argument('--epochs', type=int, default=20, metavar='N',
help='number of epochs to train (default: 20)')
parser.add_argument('--cycle', type=int, default=4, metavar='N',
help='number of epochs to train (default: 4)')
parser.add_argument('--P', type=int, default=10, help='model recording period (default: 10)')
parser.add_argument('--lr_max', type=float, default=0.05, metavar='LR1',
help='initial learning rate (default: 0.05)')
parser.add_argument('--lr_min', type=float, default=0.0005, metavar='LR2',
help='initial learning rate (default: 0.0001)')
parser.add_argument('--momentum', type=float, default=0.9, metavar='M',
help='SGD momentum (default: 0.9)')
parser.add_argument('--wd', type=float, default=5e-4, metavar='WD',
help='weight decay (default: 1e-4)')
parser.add_argument('--seed', type=int, default=1, metavar='S', help='random seed (default: 1)')
args = parser.parse_args()
use_cuda = torch.cuda.is_available()
if use_cuda:
args.device = torch.device("cuda")
else:
args.device = torch.device("cpu")
assert args.cycle % 2 == 0, 'Cycle length should be even'
os.makedirs(args.dir, exist_ok=True)
with open(os.path.join(args.dir, 'fge.sh'), 'w') as f:
f.write(' '.join(sys.argv))
f.write('\n')
torch.backends.cudnn.benchmark = True
torch.manual_seed(args.seed)
torch.cuda.manual_seed(args.seed)
print("Using model %s" % args.model)
model_cfg = getattr(models, args.model)
print("Loading dataset %s from %s" % (args.dataset, args.data_path))
loaders, num_classes = data.loaders(
args.dataset,
args.data_path,
args.batch_size,
args.num_workers,
model_cfg.transform_train,
model_cfg.transform_test,
use_validation=not args.use_test,
split_classes=args.split_classes,
)
print("Preparing model")
print(*model_cfg.args)
model = model_cfg.base(*model_cfg.args, num_classes=num_classes, **model_cfg.kwargs)
model.to(args.device)
optimizer = torch.optim.SGD(model.parameters(), lr=args.lr_max, momentum=args.momentum, weight_decay=args.wd)
num_model = args.epochs // args.P
model_list = []
model_list.append(model)
swa_n = np.zeros(int(num_model))
optimizer_list = []
optimizer_list.append(optimizer)
for i in range(int(num_model)):
model1 = model_cfg.base(*model_cfg.args, num_classes=num_classes, **model_cfg.kwargs)
model1.to(args.device)
optimizers = torch.optim.SGD(model1.parameters(), lr=args.lr_max, momentum=args.momentum, weight_decay=args.wd)
model_list.append(model1)
optimizer_list.append(optimizers)
criterion = utils.cross_entropy
checkpoint = torch.load(args.ckpt)
start_epoch = 0
model.load_state_dict(checkpoint['state_dict'])
optimizer.load_state_dict(checkpoint['optimizer'])
ensemble_size = 0
pfge_predictions_sum = np.zeros((len(loaders['test'].dataset), num_classes))
columns = ['ep', 'lr', 'tr_loss', 'tr_acc', 'te_loss', 'te_acc', 'pfge_ens_acc', 'time']
for epoch in range(args.epochs):
time_ep = time.time()
lr_schedule = utils.cyclic_learning_rate(epoch, args.cycle, args.lr_max, args.lr_min)
i = epoch // args.P
if i == 0:
train_res = utils.train_epochs(loaders['train'], model, criterion, optimizer, lr_schedule=lr_schedule,
cuda=use_cuda)
test_res = utils.eval(loaders['test'], model, criterion, cuda=use_cuda)
time_ep = time.time() - time_ep
pfge_ens_acc = None
if (epoch % args.cycle + 1) == args.cycle // 2:
utils.moving_average(model_list[i+1], model, 1.0 / (swa_n[i] + 1))
swa_n[i] += 1
if (epoch + 1) % args.P == 0:
ensemble_size += 1
utils.bn_update(loaders["train"], model_list[i+1])
pfge_res = utils.predict(loaders["test"], model_list[i+1])
pfge_predictions = pfge_res["predictions"]
targets = pfge_res["targets"]
pfge_predictions_sum += pfge_predictions
pfge_ens_acc = 100.0 * np.mean(np.argmax(pfge_predictions_sum, axis=1) == targets)
utils.save_checkpoint(args.dir, epoch + 1, name="pfge", state_dict=model_list[i+1].state_dict())
else:
train_res = utils.train_epochs(loaders['train'], model_list[i], criterion, optimizer_list[i], lr_schedule=lr_schedule, cuda=use_cuda)
test_res = utils.eval(loaders['test'], model_list[i], criterion, cuda=use_cuda)
time_ep = time.time() - time_ep
if (epoch % args.cycle + 1) == args.cycle // 2:
utils.moving_average(model_list[i+1], model_list[i], 1.0/(swa_n[i] + 1))
swa_n[i] += 1
if (epoch + 1) % args.P == 0:
ensemble_size += 1
utils.bn_update(loaders["train"], model_list[i + 1])
pfge_res = utils.predict(loaders["test"], model_list[i + 1])
pfge_predictions = pfge_res["predictions"]
targets = pfge_res["targets"]
pfge_predictions_sum += pfge_predictions
pfge_ens_acc = 100.0 * np.mean(np.argmax(pfge_predictions_sum, axis=1) == targets)
utils.save_checkpoint(args.dir, epoch + 1, name="pfge", state_dict=model_list[i + 1].state_dict())
values = [epoch + 1, lr_schedule(1.0), train_res['loss'], train_res['accuracy'], test_res['loss'],
test_res['accuracy'], pfge_ens_acc, time_ep]
table = tabulate.tabulate([values], columns, tablefmt='simple', floatfmt='9.4f')
if epoch % 40 == 0:
table = table.split('\n')
table = '\n'.join([table[1]] + table)
else:
table = table.split('\n')[2]
print(table)