-
Notifications
You must be signed in to change notification settings - Fork 0
/
train_conv.py
128 lines (111 loc) · 3.21 KB
/
train_conv.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
import time
import numpy as np
import torch as th
import torchvision.utils as tvu
from torch.utils.tensorboard.writer import SummaryWriter
import data
import models
import optim
import shltutils.filters as filters
import util
bs = 16
patch_size = 96
color = False
rotate = True
flip = True
log_freq = 20
n_w = 125
dataset = data.BSDS(color, bs, patch_size, rotate, flip)
h = th.tensor([
0.0104933261758410, -0.0263483047033631, -0.0517766952966370,
0.276348304703363, 0.582566738241592, 0.276348304703363,
-0.0517766952966369, -0.0263483047033631, 0.0104933261758408
],
device='cuda')
lamda = th.ones((21, ), dtype=th.float32, device='cuda')
lamda = lamda.float()
h0, _ = filters.dfilters('dmaxflat4', 'd')
P = th.from_numpy(filters.modulate2(h0, 'c')).float()
R = models.GMMConv(
n_scales=2,
symmetric=True,
vmin=-.5,
vmax=.5,
n_w=n_w,
w_init='student-t',
lamda=lamda.cuda(),
h=h.cuda(),
dims=(patch_size, patch_size),
P=P.cuda()
).cuda()
R.set_eta()
R.set_sigma(0)
ipalm = False
t = str(time.time())
writer = SummaryWriter(log_dir='./log/shearlets/' + t)
lrs = {
'w.w': 1e-4,
'lamda': 5e-5,
'h': 5e-5,
'P': 5e-5,
}
if ipalm:
groups = []
for k, v in R.named_parameters():
groups.append({
'params': v,
'name': k,
})
optimizer = optim.IPalm(groups, eps=1e-5)
else:
groups = []
for k, v in R.named_parameters():
# if k not in lrs.keys():
# print(k)
groups.append({
'params': v,
'lr': lrs[k],
'name': k,
})
optimizer = optim.AdaBelief(groups)
def loss_criterion(y, R, sigmas):
n = [th.randn_like(y) for _ in range(len(sigmas))]
def closure(compute_grad=False):
with th.set_grad_enabled(compute_grad):
loss_sm = 0
for sigma, noise in zip(sigmas, n):
R.set_sigma(sigma)
x = y + sigma * noise
loss_sm += ((sigma * R.grad(x)[1] - noise)**2).sum()
return [loss_sm / bs]
return closure
for i, y in enumerate(dataset):
closure = loss_criterion(y, R, np.random.rand(10) * .4)
if i % log_freq == 0:
writer.add_scalar('loss/score', sum(closure()).item(), global_step=i)
# writer.add_figure('theta', vis.vis(R), i)
for s in [0.025, 0.05, 0.1, 0.2]:
with th.no_grad():
R.set_sigma(s)
x = y[:25] + s * th.randn_like(y[:25])
y_hat = x - R.grad(x)[1] * s**2
stack = th.concat((x[:8], y_hat[:8], y[:8]), dim=0)
writer.add_scalar(
f'psnr {s:.3f}', util.psnr(y[:25], y_hat), global_step=i
)
writer.add_image(
f'test {s:.3f}',
tvu.make_grid(th.clip(stack, 0, 1), nrow=8),
global_step=i
)
R.set_sigma(0)
th.save(R.state_dict(), f'./out/shearlets/state_{i:06d}.pth')
if ipalm:
loss = optimizer.step(closure)
else:
optimizer.zero_grad()
loss = closure(True)
sum(loss).backward()
optimizer.step()
# Update eta after each parameter update
R.set_eta()