-
Notifications
You must be signed in to change notification settings - Fork 2
/
train_search.py
192 lines (151 loc) · 6.42 KB
/
train_search.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 glob
import logging
import os
import sys
import time
import numpy as np
import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.utils
import torchvision.datasets as dset
import utils as utils
from architect import Architect
from model_search import Network
from search_config import args
if args.debug:
args.save = 'log/debug-search-{}-{}'.format(args.save, time.strftime("%Y%m%d-%H%M%S"))
else:
args.save = 'log/search-{}-{}-{}'.format(args.dataset, args.save, time.strftime("%Y%m%d-%H%M%S"))
if not os.path.exists('log'):
os.mkdir('log')
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)
if args.dataset == 'cifar10':
CIFAR_CLASSES = 10
elif args.dataset == 'cifar100':
CIFAR_CLASSES = 100
else:
raise ValueError('No Defined Dataset!!!')
def main():
utils.set_seed(seed=0)
logging.info('gpu device = %d' % args.gpu_id)
logging.info("args = %s", args)
criterion = nn.CrossEntropyLoss()
criterion = criterion.cuda()
model = Network(args, args.init_channels, CIFAR_CLASSES, args.layers, criterion)
model = model.cuda()
logging.info("param size = %fMB", utils.count_parameters_in_MB(model))
optimizer = torch.optim.SGD(
model.parameters(),
args.learning_rate,
momentum=args.momentum,
weight_decay=args.weight_decay)
architect = Architect(model, args)
if args.dataset == 'cifar10':
train_transform, valid_transform = utils.data_transforms_cifar10(args)
train_data = dset.CIFAR10(root=args.data + args.dataset, train=True, download=True,
transform=train_transform)
elif args.dataset == 'cifar100':
train_transform, valid_transform = utils.data_transforms_cifar100(args)
train_data = dset.CIFAR100(root=args.data + args.dataset, train=True, download=True,
transform=train_transform)
else:
raise ValueError('No Defined Dataset!!!')
num_train = len(train_data)
indices = list(range(num_train))
split = int(np.floor(args.train_portion * num_train))
if args.debug:
split = args.batch_size
num_train = 2 * args.batch_size
train_queue = torch.utils.data.DataLoader(
train_data, batch_size=args.batch_size,
sampler=torch.utils.data.sampler.SubsetRandomSampler(indices[:split]),
pin_memory=True, num_workers=0)
valid_queue = torch.utils.data.DataLoader(
train_data, batch_size=args.batch_size,
sampler=torch.utils.data.sampler.SubsetRandomSampler(indices[split:num_train]),
pin_memory=True, num_workers=0)
scheduler = torch.optim.lr_scheduler.CosineAnnealingLR(
optimizer, float(args.epochs), eta_min=args.learning_rate_min)
for epoch in range(args.epochs):
scheduler.step()
lr = scheduler.get_lr()[0]
logging.info('epoch %d lr %e', epoch, lr)
# training
train_acc, train_obj = train(train_queue, valid_queue, model, architect, criterion, optimizer, lr, epoch)
logging.info('train_acc %f', train_acc)
# validation
valid_acc, valid_obj = infer(valid_queue, model, criterion)
logging.info('valid_acc %f', valid_acc)
# print arch params
logging.info('normal alpha = %s', F.softmax(model.alphas_normal, dim=-1))
logging.info('normal beta = %s', F.sigmoid(model.betas_normal))
logging.info('reduce alpha = %s', F.softmax(model.alphas_reduce, dim=-1))
logging.info('reduce beta = %s', F.sigmoid(model.betas_reduce))
genotype = model.genotype()
logging.info('genotype = %s', genotype)
utils.save(model, os.path.join(args.save, 'weights.pt'))
def train(train_queue, valid_queue, model, architect, criterion, optimizer, lr, epoch):
objs = utils.AverageMeter()
top1 = utils.AverageMeter()
top5 = utils.AverageMeter()
args.epoch = epoch
if args.warmup and epoch < args.warmup_epoch:
logging.info('epoch %d warming up!!!', epoch)
else:
logging.info('epoch %d train arch!!!', epoch)
for step, (input, target) in enumerate(train_queue):
args.step = step
model.train()
n = input.size(0)
input = input.cuda()
target = target.cuda(non_blocking=True)
if architect is not None:
if args.warmup and epoch < args.warmup_epoch:
pass
else:
# get a random minibatch from the search queue with replacement
input_search, target_search = next(iter(valid_queue))
input_search = input_search.cuda()
target_search = target_search.cuda(non_blocking=True)
architect.step(input, target, input_search, target_search, lr, optimizer, unrolled=args.unrolled)
optimizer.zero_grad()
logits = model(input)
loss = criterion(logits, target)
loss.backward()
nn.utils.clip_grad_norm_(model.parameters(), args.grad_clip)
optimizer.step()
prec1, prec5 = utils.accuracy(logits, target, topk=(1, 5))
objs.update(loss.item(), n)
top1.update(prec1.item(), n)
top5.update(prec5.item(), n)
if step % args.report_freq == 0:
logging.info('train %03d %e %f %f', step, objs.avg, top1.avg, top5.avg)
return top1.avg, objs.avg
def infer(valid_queue, model, criterion):
objs = utils.AverageMeter()
top1 = utils.AverageMeter()
top5 = utils.AverageMeter()
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.item(), n)
top1.update(prec1.item(), n)
top5.update(prec5.item(), n)
if step % args.report_freq == 0:
logging.info('valid %03d %e %f %f', step, objs.avg, top1.avg, top5.avg)
return top1.avg, objs.avg
if __name__ == '__main__':
utils.run_func(args, main)