-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathtrain.py
192 lines (161 loc) · 8.05 KB
/
train.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
import os
import sys
import numpy as np
import time
import torch
import utils
from tqdm import tqdm
import glob
import logging
import argparse
import torch.nn as nn
import genotypes as genotypes
import torch.utils
from torch.utils.tensorboard import SummaryWriter
import torchvision.datasets as dset
import torch.backends.cudnn as cudnn
from thop import profile
from model import NetworkCIFAR as Network
parser = argparse.ArgumentParser("training cifar-10")
parser.add_argument('--workers', type=int, default=16, help='number of workers to load dataset')
parser.add_argument('--data', type=str, default='datapath', help='location of the data corpus')
parser.add_argument('--dataset', type=str, default='cifar10', help='choose dataset')
parser.add_argument('--batch_size', type=int, default=96, help='batch size')
parser.add_argument('--learning_rate', type=float, default=0.025, help='init learning rate')
parser.add_argument('--momentum', type=float, default=0.9, help='momentum')
parser.add_argument('--weight_decay', type=float, default=3e-4, help='weight decay')
parser.add_argument('--gpu', type=int, default=0, help='gpu device id')
parser.add_argument('--epochs', type=int, default=600, help='num of training epochs')
parser.add_argument('--init_channels', type=int, default=36, help='num of init channels')
parser.add_argument('--layers', type=int, default=20, help='total number of layers')
parser.add_argument('--model_path', type=str, default='saved_models', help='path to save the model')
parser.add_argument('--auxiliary', action='store_true', default=False, help='use auxiliary tower')
parser.add_argument('--auxiliary_weight', type=float, default=0.4, help='weight for auxiliary loss')
parser.add_argument('--cutout', action='store_true', default=False, help='use cutout')
parser.add_argument('--cutout_length', type=int, default=16, help='cutout length')
parser.add_argument('--drop_path_prob', type=float, default=0.3, help='drop path probability')
parser.add_argument('--save', type=str, default='exp', help='experiment name')
parser.add_argument('--seed', type=int, default=0, help='random seed')
parser.add_argument('--arch', type=str, default='DrNAS_cifar10', help='which architecture to use')
parser.add_argument('--grad_clip', type=float, default=5, help='gradient clipping')
args = parser.parse_args()
# set up logs
args.save = './experiments/{}/eval-{}-{}-{}-{}'.format(
args.dataset, args.save, time.strftime("%Y%m%d-%H%M%S"), args.arch, args.seed)
if args.cutout:
args.save += '-cutout-' + str(args.cutout_length)
if args.auxiliary:
args.save += '-auxiliary-' + str(args.auxiliary_weight)
utils.create_exp_dir(args.save, scripts_to_save=glob.glob('*.py'))
log_format = '%(asctime)s %(message)s'
logging.basicConfig(stream=sys.stdout, level=logging.INFO, format=log_format, datefmt='%m/%d %I:%M:%S %p')
fh = logging.FileHandler(os.path.join(args.save, 'log.txt'))
fh.setFormatter(logging.Formatter(log_format))
logging.getLogger().addHandler(fh)
writer = SummaryWriter(args.save)
CIFAR_CLASSES = 10
if args.dataset=='cifar100':
CIFAR_CLASSES = 100
def main():
if not torch.cuda.is_available():
logging.info('No GPU device available')
sys.exit(1)
np.random.seed(args.seed)
torch.cuda.set_device(args.gpu)
cudnn.benchmark = True
torch.manual_seed(args.seed)
cudnn.enabled = True
torch.cuda.manual_seed(args.seed)
logging.info('gpu device = %d' % args.gpu)
logging.info("args = %s", args)
genotype = eval("genotypes.%s" % args.arch)
print('---------Genotype---------')
logging.info(genotype)
print('--------------------------')
model = Network(args.init_channels, CIFAR_CLASSES, args.layers, args.auxiliary, genotype)
model.drop_path_prob = 0
macs, params = profile(model, inputs=(torch.randn(1, 3, 32, 32), ), verbose=False)
logging.info("param = %f, flops = %f", params, macs)
model = model.cuda()
logging.info("param size = %fMB", utils.count_parameters_in_MB(model))
criterion = nn.CrossEntropyLoss()
criterion = criterion.cuda()
optimizer = torch.optim.SGD(
model.parameters(),
args.learning_rate,
momentum=args.momentum,
weight_decay=args.weight_decay
)
train_transform, valid_transform = utils._data_transforms_cifar10(args)
if args.dataset == 'cifar100':
train_data = dset.CIFAR100(root=args.data, train=True, download=True, transform=train_transform)
valid_data = dset.CIFAR100(root=args.data, train=False, download=True, transform=valid_transform)
else:
train_data = dset.CIFAR10(root=args.data, train=True, download=True, transform=train_transform)
valid_data = dset.CIFAR10(root=args.data, train=False, download=True, transform=valid_transform)
train_queue = torch.utils.data.DataLoader(train_data, batch_size=args.batch_size, shuffle=True, pin_memory=True)
valid_queue = torch.utils.data.DataLoader(valid_data, batch_size=args.batch_size, shuffle=False, pin_memory=True)
scheduler = torch.optim.lr_scheduler.CosineAnnealingLR(optimizer, float(args.epochs))
best_acc = train_acc = valid_acc = 0.0
epoch_bar = tqdm(range(args.epochs), position=0, leave=True)
for epoch in epoch_bar:
# logging.info('epoch %d lr %e', epoch, scheduler.get_last_lr()[0])
model.drop_path_prob = args.drop_path_prob * epoch / args.epochs
description = 'Epoch [{}/{}] | LR:{} | Train:{} | Validation:{} | Best: {}'.format(epoch+1, args.epochs, scheduler.get_last_lr()[0], train_acc, valid_acc, best_acc)
train_acc, train_obj = train(train_queue, model, criterion, optimizer)
# logging.info('train_acc %f', train_acc)
description = 'Epoch [{}/{}] | LR:{} | Train:{} | Validation:{} | Best: {}'.format(epoch+1, args.epochs, scheduler.get_last_lr()[0], train_acc, valid_acc, best_acc)
epoch_bar.set_description(description)
valid_acc, valid_obj = infer(valid_queue, model, criterion)
if valid_acc > best_acc:
best_acc = valid_acc
# logging.info('valid_acc %f, best_acc %f', valid_acc, best_acc)
description = 'Epoch [{}/{}] | LR:{} | Train:{} | Validation:{} | Best: {}'.format(epoch+1, args.epochs, scheduler.get_last_lr()[0], train_acc, valid_acc, best_acc)
epoch_bar.set_description(description)
writer.add_scalar("acc/train", train_acc, epoch)
writer.add_scalar("acc/valid_best", best_acc, epoch)
writer.add_scalar("acc/valid", valid_acc, epoch)
scheduler.step()
utils.save(model, os.path.join(args.save, 'weights.pt'))
def train(train_queue, model, criterion, optimizer):
objs = utils.AvgrageMeter()
top1 = utils.AvgrageMeter()
top5 = utils.AvgrageMeter()
model.train()
for step, (input, target) in enumerate(train_queue):
input = input.cuda(non_blocking=True)
target = target.cuda(non_blocking=True)
optimizer.zero_grad()
logits, logits_aux = model(input)
loss = criterion(logits, target)
if args.auxiliary:
loss_aux = criterion(logits_aux, target)
loss += args.auxiliary_weight*loss_aux
loss.backward()
nn.utils.clip_grad_norm_(model.parameters(), args.grad_clip)
optimizer.step()
prec1, prec5 = utils.accuracy(logits, target, topk=(1, 5))
n = input.size(0)
objs.update(loss.data, n)
top1.update(prec1.data, n)
top5.update(prec5.data, n)
return top1.avg, objs.avg
def infer(valid_queue, model, criterion):
objs = utils.AvgrageMeter()
top1 = utils.AvgrageMeter()
top5 = utils.AvgrageMeter()
model.eval()
with torch.no_grad():
for step, (input, target) in enumerate(valid_queue):
input = input.cuda()
target = target.cuda(non_blocking=True)
logits, _ = model(input)
loss = criterion(logits, target)
prec1, prec5 = utils.accuracy(logits, target, topk=(1, 5))
n = input.size(0)
objs.update(loss.data, n)
top1.update(prec1.data, n)
top5.update(prec5.data, n)
return top1.avg, objs.avg
if __name__ == '__main__':
main()