-
Notifications
You must be signed in to change notification settings - Fork 0
/
glo_main.py
155 lines (125 loc) · 4.88 KB
/
glo_main.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
import torch
import torchvision
import os
from torch import optim
from model import Generator2
from args import get_opt, print_opt
from util import setup, to_variable, denorm, save_checkpoint, get_lastest_ckpt
import numpy as np
from data_loader import get_loader, process_z
from laplacian_loss import laplacian_loss
opt = get_opt()
print_opt()
# hyper parameters
n_epochs = opt.epochs
batch_size = opt.batch_size
z_dim = opt.z_dim
x_dim = opt.x_dim
sample_size = opt.sample_size
lr = opt.lr
log_step = 100
sample_step = 1000
sample_path, model_path = setup()
image_path = os.path.join(os.getcwd(), 'CelebA', '128_crop')
train_loader = get_loader(image_path=image_path,
image_size=opt.x_dim,
batch_size=opt.batch_size,
num_workers=2)
image_path = os.path.join(os.getcwd(), 'CelebA', '128_crop')
img_size = opt.x_dim
# choose generator type laplacian generator or common dcgan generator
generator = Generator2()
if opt.gpu:
generator.cuda()
if opt.optim == 'Adam':
g_optimizer = optim.Adam(generator.parameters(), lr, betas=(opt.beta1, opt.beta2))
elif opt.optim == 'SGD':
g_optimizer = optim.SGD(generator.parameters(), lr)
else:
print('optimizer is not set correctly. Adam will be used')
g_optimizer = optim.Adam(generator.parameters(), lr, betas=(opt.beta1, opt.beta2))
z_file_name = 'processed_z'
z_file_name += '_b' + str(opt.batch_size)
z_file_name += '_z' + str(opt.z_dim)
z_file_name += '.pt'
if not os.path.isfile(z_file_name):
print("no processed z. so create one..")
process_z(opt.x_dim, opt.z_dim, opt.batch_size, './', './')
if opt.resume == 'auto':
# auto option load the checkpoint of the lastest models of the lastest epoch
ckpt_path = get_lastest_ckpt()
print("=> ", ckpt_path, " is being loaded")
ckpt = torch.load(ckpt_path)
opt.start_epoch = ckpt['epoch']
generator.load_state_dict(ckpt['state_dict'])
z_in_ball = ckpt['latent']
g_optimizer.load_state_dict(ckpt['optimizer'])
elif opt.resume is not None:
if os.path.isfile(opt.resume):
ckpt = torch.load(opt.resume)
opt.start_epoch = ckpt['epoch']
generator.load_state_dict(ckpt['state_dict'])
z_in_ball = ckpt['latent']
g_optimizer.load_state_dict(ckpt['optimizer'])
else:
print("No check point file in input path")
else:
z_in_ball = torch.load(z_file_name)
if opt.gpu:
learnable_z = z_in_ball.cuda()
else:
learnable_z = z_in_ball
total_step = len(train_loader)
for epoch in range(opt.start_epoch, n_epochs):
for i, x in enumerate(train_loader):
if i == 4880:
break
x = to_variable(x)
z = to_variable(learnable_z[i], requires_grad=True)
x_hat = generator.forward(z)
l1_loss = opt.l1_weight * torch.mean(torch.abs(x - x_hat))
lap_loss = laplacian_loss(x, x_hat, n_levels=-1, cuda=opt.gpu)
loss = l1_loss + lap_loss
g_optimizer.zero_grad()
loss.backward()
g_optimizer.step()
if opt.gpu:
grad = z.grad.data.cuda()
else:
grad = z.grad.data
z_update = learnable_z[i] - opt.alpha * grad
z_update = z_update.cpu().numpy()
norm = np.sqrt(np.sum(z_update ** 2, axis=1))
z_update_norm = z_update / norm[:, np.newaxis]
if opt.gpu:
learnable_z[i] = torch.from_numpy(z_update_norm).cuda()
else:
learnable_z[i] = torch.from_numpy(z_update_norm).cpu()
if (i + 1) % log_step == 0:
print('Epoch [%d/%d], Step[%d/%d], loss: %f, l1: %f, lap: %f'
% (epoch + 1, n_epochs, i + 1, total_step, loss.data[0], l1_loss.data[0],
lap_loss.data[0]
))
# save the real images
if (i + 1) == sample_step:
torchvision.utils.save_image(denorm(x.data),
os.path.join(sample_path,
'real_samples-%d-%d.png' % (
epoch + 1, i + 1)), nrow=4)
# save the generated images
if (i + 1) % sample_step == 0:
torchvision.utils.save_image(denorm(x_hat.data),
os.path.join(sample_path,
'fake_samples-%d-%d.png' % (
epoch + 1, i + 1)), nrow=4)
if (epoch + 1) % opt.ckpt_step == 0:
print("saving checkpoint ..")
checkpoint_path = os.path.join(model_path, 'checkpoint_%d.pth.tar' % (epoch + 1))
save_checkpoint({
'epoch': epoch + 1,
'state_dict': generator.state_dict(),
'latent': learnable_z,
'optimizer': g_optimizer.state_dict(),
'args': opt
}, filename=checkpoint_path)
print("done.")