-
Notifications
You must be signed in to change notification settings - Fork 2
/
Copy pathtrain_search.py
205 lines (170 loc) · 9.47 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
193
194
195
196
197
198
199
200
201
202
203
204
205
import os
import random
import sys
import time
import glob
import numpy as np
import torch
from PIL import Image
import utils
import logging
import argparse
import torch.nn as nn
import torch.utils
import torch.nn.functional as F
import torch.backends.cudnn as cudnn
from model_search import Encoder, Decoder
from architect import Architect
import pytorch_msssim
import torchvision.transforms as transforms
parser = argparse.ArgumentParser("untitled")
parser.add_argument('--batch_size', type=int, default=4, help='batch size') # 64改成了4
parser.add_argument('--learning_rate', type=float, default=1e-5, help='init learning rate')
parser.add_argument('--learning_rate_min', type=float, default=1e-6, help='min 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('--report_freq', type=float, default=50, help='report frequency')
parser.add_argument('--gpu', type=int, default=0, help='gpu device id')
parser.add_argument('--epochs', type=int, default=50, help='num of training epochs')
parser.add_argument('--init_channels', type=int, default=16, help='num of init channels')
parser.add_argument('--layers', type=int, default=2, help='total number of layers')
parser.add_argument('--model_path', type=str, default='saved_models', help='path to save the model')
parser.add_argument('--save', type=str, default='EXP', help='experiment name')
parser.add_argument('--seed', type=int, default=2, help='random seed')
parser.add_argument('--grad_clip', type=float, default=5, help='gradient clipping')
parser.add_argument('--unrolled', action='store_true', default=False, help='use one-step unrolled validation loss')
parser.add_argument('--arch_learning_rate', type=float, default=4e-4, help='learning rate for arch encoding') # 3e-4
parser.add_argument('--arch_weight_decay', type=float, default=1e-3, help='weight decay for arch encoding') # 1e-3
parser.add_argument('--dataset1', type=str, default=r'C:\Users\ADMIN\Desktop\search_U2F\data_vis_ir64_160\crop_infrared160', help='Infrared images for training')
parser.add_argument('--dataset2', type=str, default=r'C:\Users\ADMIN\Desktop\search_U2F\data_vis_ir64_160\crop_visible160', help='Visible images for training')
args = parser.parse_args()
args.save = 'search-C{}-B{}-{}'.format(args.init_channels, args.batch_size, time.strftime("%Y%m%d-%H%M%S"))
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)
os.mkdir(args.save+'/output')
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 # 加速
torch.manual_seed(args.seed) # 为CUP设置随机种子
cudnn.enabled = True # 使用非确定性算法优化运行
torch.cuda.manual_seed(args.seed) # 为GPU设置随机种子
logging.info('gpu device = %d' % args.gpu)
logging.info("args = %s", args)
mse_loss = torch.nn.MSELoss().cuda()
ssim_loss = pytorch_msssim.msssim
encoder1 = Encoder(args.init_channels, args.layers).cuda()
encoder2 = Encoder(args.init_channels, args.layers).cuda()
decoder = Decoder(args.init_channels, args.layers).cuda()
# logging.info("param size = %fMB", utils.count_parameters_in_MB(model))
para1 = [{'params': encoder1.parameters(), 'lr': args.learning_rate},
{'params': decoder.parameters(), 'lr': args.learning_rate}]
para2 = [{'params': encoder2.parameters(), 'lr': args.learning_rate},
{'params': decoder.parameters(), 'lr': args.learning_rate}]
optimizer1 = torch.optim.SGD(
para1,
args.learning_rate,
momentum=args.momentum,
weight_decay=args.weight_decay)
optimizer2 = torch.optim.SGD(
para2,
args.learning_rate,
momentum=args.momentum,
weight_decay=args.weight_decay)
epochs = args.epochs
# 加载数据集
Infrared_path_list = utils.list_images(args.dataset1)
Visible_path_list = utils.list_images(args.dataset2)
# train_num = len(total_path_list)//2
random.shuffle(Infrared_path_list)
random.shuffle(Visible_path_list)
# train_num = 15000
train_num = 15000
Infrared_path_list = Infrared_path_list[:train_num]
Visible_path_list = Visible_path_list[:train_num]
train_queue1, batches = utils.load_dataset(Infrared_path_list[:train_num//2], args.batch_size)
valid_queue1, batches = utils.load_dataset(Infrared_path_list[train_num//2:train_num], args.batch_size)
train_queue2, batches = utils.load_dataset(Visible_path_list[:train_num//2], args.batch_size)
valid_queue2, batches = utils.load_dataset(Visible_path_list[train_num//2:train_num], args.batch_size)
scheduler1 = torch.optim.lr_scheduler.CosineAnnealingLR(
optimizer1, args.epochs, eta_min=args.learning_rate_min)
scheduler2 = torch.optim.lr_scheduler.CosineAnnealingLR(
optimizer2, args.epochs, eta_min=args.learning_rate_min)
architect1 = Architect(encoder1, decoder, args, mse_loss, ssim_loss)
architect2 = Architect(encoder2, decoder, args, mse_loss, ssim_loss)
train_queue = [train_queue1, train_queue2]
valid_queue = [valid_queue1, valid_queue2]
encoder = [encoder1, encoder2]
optimizer = [optimizer1, optimizer2]
architect = [architect1, architect2]
for epoch in range(epochs):
# lr = scheduler.get_lr()[0]
lr1 = scheduler1.get_last_lr()
lr2 = scheduler2.get_last_lr()
logging.info('epoch %d lr1 %e lr2 %e', epoch, lr1[0], lr2[0])
genotype_en1 = encoder1.genotype()
genotype_en2 = encoder2.genotype()
genotype_de = decoder.genotype()
logging.info('genotype_en1 = %s', genotype_en1)
logging.info('genotype_en2 = %s', genotype_en2)
logging.info('genotype_de = %s', genotype_de)
print(F.softmax(encoder1.alphas, dim=-1))
print(F.softmax(encoder2.alphas, dim=-1))
print(F.softmax(decoder.alphas, dim=-1))
logging.info('en1 = %s', F.softmax(encoder1.alphas, dim=-1))
logging.info('en2 = %s', F.softmax(encoder2.alphas, dim=-1))
logging.info('de = %s', F.softmax(decoder.alphas, dim=-1))
# training
train(train_queue, valid_queue, batches, encoder, decoder, architect, mse_loss, ssim_loss, optimizer, [lr1, lr2], epoch)
if (epoch+1) % 5 == 0:
utils.save(encoder1, os.path.join(args.save, 'encoder1_epoch'+str(epoch+1)+'.pt'))
utils.save(encoder2, os.path.join(args.save, 'encoder2_epoch'+str(epoch+1)+'.pt'))
utils.save(decoder, os.path.join(args.save, 'decoder_epoch'+str(epoch+1)+'.pt'))
scheduler1.step()
scheduler2.step()
tensor_to_pil = transforms.ToPILImage()
def train(train_queue12, valid_queue12, batches, encoder12, decoder, architect12, mse_loss, ssim, optimizer12, lr12, epoch):
encoder12[0].train()
encoder12[1].train()
decoder.train()
for batch in range(batches):
for i, train_queue, valid_queue, architect, optimizer, encoder, lr in zip(range(2), train_queue12, valid_queue12, architect12, optimizer12, encoder12, lr12):
image_paths_train = train_queue[batch * args.batch_size:(batch * args.batch_size + args.batch_size)] # 训练一批
inputs = utils.get_train_images_auto(image_paths_train).cuda() # 取出一批图片并且变成张量
image_paths_valid = valid_queue[batch * args.batch_size:(batch * args.batch_size + args.batch_size)] # 训练一批
inputs_search = utils.get_train_images_auto(image_paths_valid).cuda() # 取出一批图片并且变成张量
architect.step(inputs, inputs_search, lr, optimizer, unrolled=args.unrolled)
print(F.softmax(encoder.alphas, dim=-1))
print(F.softmax(decoder.alphas, dim=-1))
en1, en2 = encoder(inputs)
outputs = decoder(en1, en2)
optimizer.zero_grad()
ssim_loss_value = 0.
pixel_loss_value = 0.
for output, input in zip(outputs, inputs):
output, input = torch.unsqueeze(output, 0), torch.unsqueeze(input, 0)
pixel_loss_temp = mse_loss(input, output)
ssim_loss_temp = ssim(input, output, normalize=True, val_range=255)
ssim_loss_value += (1 - ssim_loss_temp)
pixel_loss_value += pixel_loss_temp
ssim_loss_value /= len(outputs)
pixel_loss_value /= len(outputs)
total_loss = pixel_loss_value + 100*ssim_loss_value
total_loss.backward()
nn.utils.clip_grad_value_(encoder.parameters(), args.grad_clip)
nn.utils.clip_grad_value_(decoder.parameters(), args.grad_clip)
optimizer.step()
if i==0:
logging.info("Infrared_epoch: %d batch: %d loss: %f", epoch, batch+1, total_loss)
else:
logging.info("Visible_epoch: %d batch: %d loss: %f", epoch, batch+1, total_loss)
if __name__ == '__main__':
main()