-
Notifications
You must be signed in to change notification settings - Fork 113
/
imagenet.py
348 lines (294 loc) · 12.7 KB
/
imagenet.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
from __future__ import print_function
import argparse
import os
import shutil
import time
import random
import torch
import torch.nn as nn
import torch.nn.parallel
import torch.backends.cudnn as cudnn
import torch.optim as optim
import torchvision.transforms as transforms
import torchvision.datasets as datasets
import torchvision.models as models
import ibnnet as customized_models
from utils import Bar, AverageMeter, accuracy, mkdir_p
# Models
default_model_names = sorted(name for name in models.__dict__
if name.islower() and not name.startswith("__")
and callable(models.__dict__[name]))
customized_models_names = sorted(name for name in customized_models.__dict__
if name.islower() and not name.startswith("__")
and callable(customized_models.__dict__[name]))
for name in customized_models.__dict__:
if name.islower() and not name.startswith("__") and callable(customized_models.__dict__[name]):
models.__dict__[name] = customized_models.__dict__[name]
model_names = default_model_names + customized_models_names
# Parse arguments
parser = argparse.ArgumentParser(description='PyTorch ImageNet Training')
# Datasets
parser.add_argument('-d', '--data', default='path to dataset', type=str)
parser.add_argument('-j', '--workers', default=4, type=int, metavar='N',
help='number of data loading workers (default: 4)')
# Optimization options
parser.add_argument('--epochs', default=100, type=int, metavar='N',
help='number of total epochs to run')
parser.add_argument('--start-epoch', default=0, type=int, metavar='N',
help='manual epoch number (useful on restarts)')
parser.add_argument('--train-batch', default=256, type=int, metavar='N',
help='train batchsize (default: 256)')
parser.add_argument('--test-batch', default=100, type=int, metavar='N',
help='test batchsize (default: 100)')
parser.add_argument('--lr', '--learning-rate', default=0.1, type=float,
metavar='LR', help='initial learning rate')
parser.add_argument('--drop', '--dropout', default=0, type=float,
metavar='Dropout', help='Dropout ratio')
parser.add_argument('--schedule', type=int, nargs='+', default=[30, 60, 90],
help='Decrease learning rate at these epochs.')
parser.add_argument('--gamma', type=float, default=0.1, help='LR is multiplied by gamma on schedule.')
parser.add_argument('--momentum', default=0.9, type=float, metavar='M',
help='momentum')
parser.add_argument('--weight-decay', '--wd', default=1e-4, type=float,
metavar='W', help='weight decay (default: 1e-4)')
# Checkpoints
parser.add_argument('-c', '--checkpoint', default='checkpoint', type=str, metavar='PATH',
help='path to save checkpoint (default: checkpoint)')
parser.add_argument('--resume', default='', type=str, metavar='PATH',
help='path to latest checkpoint (default: none)')
# Architecture
parser.add_argument('--arch', '-a', metavar='ARCH', default='resnet50',
choices=model_names,
help='model architecture: ' +
' | '.join(model_names) +
' (default: resnet50)')
parser.add_argument('--depth', type=int, default=29, help='Model depth.')
parser.add_argument('--cardinality', type=int, default=32, help='ResNet cardinality (group).')
parser.add_argument('--base-width', type=int, default=4, help='ResNet base width.')
parser.add_argument('--widen-factor', type=int, default=4,
help='Widen factor. 4 -> 64, 8 -> 128, ...')
# Miscs
parser.add_argument('--manualSeed', type=int, help='manual seed')
parser.add_argument('-e', '--evaluate', dest='evaluate', action='store_true',
help='evaluate model on validation set')
parser.add_argument('--pretrained', dest='pretrained', action='store_true',
help='use pre-trained model')
parser.add_argument('--model_weight', dest='model_weight', default=None, type=str,
help='custom pretrained model weight')
# Device options
parser.add_argument('--cpu', dest='cpu', action='store_true',
help='use cpu mode')
parser.add_argument('--gpu_id', default='1', type=str,
help='id(s) for CUDA_VISIBLE_DEVICES')
args = parser.parse_args()
state = {k: v for k, v in args._get_kwargs()}
# Use CUDA
if args.cpu:
print('Use CPU mode')
use_cuda = False
pin_memory = False
else:
os.environ['CUDA_VISIBLE_DEVICES'] = args.gpu_id
use_cuda = torch.cuda.is_available()
pin_memory = True
# Random seed
if args.manualSeed is None:
args.manualSeed = random.randint(1, 10000)
random.seed(args.manualSeed)
torch.manual_seed(args.manualSeed)
if use_cuda:
torch.cuda.manual_seed_all(args.manualSeed)
best_acc = 0 # best test accuracy
def main():
global best_acc
start_epoch = args.start_epoch # start from epoch 0 or last checkpoint epoch
if not os.path.isdir(args.checkpoint):
mkdir_p(args.checkpoint)
# Data loading code
traindir = os.path.join(args.data, 'train')
valdir = os.path.join(args.data, 'val')
normalize = transforms.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225])
train_loader = torch.utils.data.DataLoader(
datasets.ImageFolder(traindir, transforms.Compose([
transforms.RandomSizedCrop(224),
transforms.RandomHorizontalFlip(),
transforms.ToTensor(),
normalize,
])),
batch_size=args.train_batch, shuffle=True,
num_workers=args.workers, pin_memory=pin_memory)
val_loader = torch.utils.data.DataLoader(
datasets.ImageFolder(valdir, transforms.Compose([
transforms.Scale(256),
transforms.CenterCrop(224),
transforms.ToTensor(),
normalize,
])),
batch_size=args.test_batch, shuffle=False,
num_workers=args.workers, pin_memory=pin_memory)
# create model
print("=> creating model '{}'".format(args.arch))
if args.arch.startswith('resnext'):
model = models.__dict__[args.arch](
pretrained=args.pretrained,
baseWidth=args.base_width,
cardinality=args.cardinality,
)
else:
model = models.__dict__[args.arch](pretrained=args.pretrained)
if use_cuda:
model = torch.nn.DataParallel(model).cuda()
cudnn.benchmark = True
print(model)
if args.model_weight:
model_weight = torch.load(args.model_weight)
model.load_state_dict(model_weight['state_dict'], strict=False)
print(' Total params: %.2fM' % (sum(p.numel() for p in model.parameters())/1000000.0))
# define loss function (criterion) and optimizer
criterion = nn.CrossEntropyLoss()
if use_cuda:
criterion = criterion.cuda()
optimizer = optim.SGD(model.parameters(),
lr=args.lr,
momentum=args.momentum,
weight_decay=args.weight_decay)
# Resume
if args.resume:
# Load checkpoint.
print('==> Resuming from checkpoint..')
assert os.path.isfile(args.resume), 'Error: no checkpoint directory found!'
args.checkpoint = os.path.dirname(args.resume)
checkpoint = torch.load(args.resume)
best_acc = checkpoint['best_acc']
start_epoch = checkpoint['epoch']
model.load_state_dict(checkpoint['state_dict'])
optimizer.load_state_dict(checkpoint['optimizer'])
if args.evaluate:
print('\nEvaluation only')
test_loss, test_acc1, test_acc5 = test(val_loader, model, criterion, start_epoch, use_cuda)
print(' Test Loss: %.8f, Top1 Acc: %.2f, Top5 Acc: %.2f'
% (test_loss, test_acc1, test_acc5))
print(' Top1 Err: %.2f, Top5 Err: %.2f' % (100.0 - test_acc1, 100.0 - test_acc5))
return
# Train and val
for epoch in range(start_epoch, args.epochs):
adjust_learning_rate(optimizer, epoch)
print('\nEpoch: [%d | %d] LR: %f' % (epoch + 1, args.epochs, state['lr']))
train_loss, train_acc = train(train_loader, model, criterion, optimizer, epoch, use_cuda)
test_loss, test_acc1, test_acc = test(val_loader, model, criterion, epoch, use_cuda)
# save model
is_best = test_acc > best_acc
best_acc = max(test_acc, best_acc)
save_checkpoint({
'epoch': epoch + 1,
'state_dict': model.state_dict(),
'acc': test_acc,
'best_acc': best_acc,
'optimizer': optimizer.state_dict(),
}, is_best, checkpoint=args.checkpoint)
print('Best acc:')
print(best_acc)
def train(train_loader, model, criterion, optimizer, epoch, use_cuda):
# switch to train mode
model.train()
batch_time = AverageMeter()
data_time = AverageMeter()
losses = AverageMeter()
top1 = AverageMeter()
top5 = AverageMeter()
end = time.time()
bar = Bar('P', max=len(train_loader))
for batch_idx, (inputs, targets) in enumerate(train_loader):
# measure data loading time
data_time.update(time.time() - end)
if use_cuda:
inputs, targets = inputs.cuda(), targets.cuda(non_blocking=True)
# compute output
outputs = model(inputs)
loss = criterion(outputs, targets)
prec1, prec5 = accuracy(outputs.data, targets.data, topk=(1, 5))
# measure accuracy and record loss
losses.update(loss.item(), inputs.size(0))
top1.update(prec1.item(), inputs.size(0))
top5.update(prec5.item(), inputs.size(0))
# compute gradient and do SGD step
optimizer.zero_grad()
loss.backward()
optimizer.step()
# measure elapsed time
batch_time.update(time.time() - end)
end = time.time()
# plot progress
if (batch_idx+1) % 10 == 0:
print('({batch}/{size}) D: {data:.2f}s | B: {bt:.2f}s | T: {total:} | '
'E: {eta:} | L: {loss:.3f} | t1: {top1: .3f} | t5: {top5: .3f}'.format(
batch=batch_idx + 1,
size=len(train_loader),
data=data_time.val,
bt=batch_time.val,
total=bar.elapsed_td,
eta=bar.eta_td,
loss=losses.avg,
top1=top1.avg,
top5=top5.avg,
))
bar.next()
bar.finish()
return (losses.avg, top5.avg)
def test(val_loader, model, criterion, epoch, use_cuda):
global best_acc
batch_time = AverageMeter()
data_time = AverageMeter()
losses = AverageMeter()
top1 = AverageMeter()
top5 = AverageMeter()
# switch to evaluate mode
model.eval()
end = time.time()
bar = Bar('P', max=len(val_loader))
for batch_idx, (inputs, targets) in enumerate(val_loader):
# measure data loading time
data_time.update(time.time() - end)
if use_cuda:
inputs, targets = inputs.cuda(), targets.cuda()
# compute output
end = time.time()
outputs = model(inputs)
batch_time.update(time.time() - end)
loss = criterion(outputs, targets)
prec1, prec5 = accuracy(outputs.data, targets.data, topk=(1, 5))
# measure accuracy and record loss
losses.update(loss.item(), inputs.size(0))
top1.update(prec1.item(), inputs.size(0))
top5.update(prec5.item(), inputs.size(0))
# plot progress
if (batch_idx+1) % 10 == 0:
print('({batch}/{size}) D: {data:.2f}s | B: {bt:.2f}s | T: {total:} | '
'E: {eta:} | L: {loss:.3f} | t1: {top1: .3f} | t5: {top5: .3f}'.format(
batch=batch_idx + 1,
size=len(val_loader),
data=data_time.avg,
bt=batch_time.avg,
total=bar.elapsed_td,
eta=bar.eta_td,
loss=losses.avg,
top1=top1.avg,
top5=top5.avg,
))
bar.next()
bar.finish()
return (losses.avg, top1.avg, top5.avg)
def save_checkpoint(state, is_best, checkpoint='checkpoint', filename='checkpoint.pth'):
filepath = os.path.join(checkpoint, filename)
torch.save(state, filepath)
if is_best:
shutil.copyfile(filepath, os.path.join(checkpoint, 'model_best.pth'))
def adjust_learning_rate(optimizer, epoch):
global state
if epoch in args.schedule:
state['lr'] *= args.gamma
for param_group in optimizer.param_groups:
param_group['lr'] = state['lr']
if __name__ == '__main__':
main()