-
Notifications
You must be signed in to change notification settings - Fork 2
/
Copy pathtree_energy_test.py
190 lines (152 loc) · 7.85 KB
/
tree_energy_test.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
import Tree
import math
import torch
from torch import nn
import torchvision
import numpy as np
import pickle
import utilities
from torch.utils.data import DataLoader
import torch.nn.functional as F
import matplotlib.pyplot as plt
from torchvision.utils import make_grid
from torch.utils.data import Subset
import ssl
ssl._create_default_https_context = ssl._create_unverified_context
#Reduces data to specified number of examples per category
def get_data(shuf=False, data=2):
transform = torchvision.transforms.Compose([torchvision.transforms.ToTensor()])
if data == 0:
trainset = torchvision.datasets.MNIST(root='./data', train=True, download=False, transform=transform)
elif data == 1:
trainset = torchvision.datasets.FashionMNIST(root='./data', train=True, download=False, transform=transform)
elif data == 2:
trainset = torchvision.datasets.CIFAR10(root='./data', train=True, download=False, transform=transform)
testset = torchvision.datasets.CIFAR10(root='./data', train=False, download=False, transform=transform)
elif data == 3:
trainset = torchvision.datasets.SVHN(root='./data', split='train', download=False, transform=transform)
elif data == 4:
dir = 'C:/Users/nalon/Documents/PythonScripts/DataSets/tiny-imagenet-200/train'
trainset = torchvision.datasets.ImageFolder(dir, transform=transform)
dirt = 'C:/Users/nalon/Documents/PythonScripts/DataSets/tiny-imagenet-200/test'
testset = torchvision.datasets.ImageFolder(dirt, transform=transform)
else:
transform = torchvision.transforms.Compose([torchvision.transforms.ToTensor(),
torchvision.transforms.CenterCrop(128)])
dir = 'C:/Users/nalon/Documents/PythonScripts/DataSets/CalTech256'
trainset = torchvision.datasets.ImageFolder(dir, transform=transform)
train_loader = torch.utils.data.DataLoader(trainset, batch_size=1, shuffle=shuf)
test_loader = torch.utils.data.DataLoader(testset, batch_size=10000, shuffle=True)
return train_loader, test_loader
def train_online(arch=2, data=4, dev='cuda', max_iter=5000, wtupType=0, num_seeds=10, alpha=10, in_dim=28, in_chn=1, chnls=200,
shuf=True, b_sim=2, simf=1):
with torch.no_grad():
tn_energies = []
tst_energies = []
for d in range(4):
tn_energies.append(torch.zeros(num_seeds, 3))
tst_energies.append(torch.zeros(num_seeds, 3))
# Memorize
for s in range(num_seeds):
model = Tree.Tree(in_dim=in_dim, in_chnls=in_chn, wt_up=wtupType, chnls=chnls, alpha=alpha, arch=arch,
simFunc=simf, b_sim=b_sim, lmbd=.5).to(dev)
train_loader, test_loader = get_data(shuf=shuf, data=data)
mem_images = torch.zeros(0, in_chn, in_dim, in_dim).to(dev)
#Train Network
for batch_idx, (images, y) in enumerate(train_loader):
images = images.to(dev)
mem_images = torch.cat((mem_images, images), dim=0)
model.update_wts(images)
if batch_idx % 100 == 0:
print(batch_idx)
if batch_idx >= max_iter:
break
#Get Energy with different inference processes
get_train_energies(model, mem_images, tn_energies, s)
get_test_energies(model, test_loader, tst_energies, s)
print(f'\nSeed:{s}\n'
f'Max:{tn_energies[0][s]}, {tst_energies[0][s]}\n '
f'ArgMax:{tn_energies[1][s][0:2]}, {tst_energies[1][s][0:2]}\n'
f'Softmax:{tn_energies[2][s][0:2]}, {tst_energies[2][s][0:2]}\n '
f'Rand:{tn_energies[3][s][0:2]}, {tst_energies[3][s][0:2]}')
with open(f'data/AA_Online_TreeEnergy_Arch{arch}_max_iter{max_iter}.data', 'wb') as filehandle:
pickle.dump([tn_energies, tst_energies], filehandle)
def corrupt_test(noise, noise_tp, mem_unit, mem_images, rec_thr=.01, rec_type=0):
with torch.no_grad():
num_img = mem_images.size(0)
num_batch = int(num_img / 500) + 1
mse = 0
recalled = 0
mse_n = 0
recalled_n = 0
mse_msk = 0
recalled_msk = 0
for b in range(num_batch):
img = mem_images[int(b*500):int(b*500) + 500].clone()
if noise_tp == 0:
imgn = torch.clamp(img + torch.randn_like(img) * noise, min=0, max=1)
else:
imgn = img + torch.randn_like(img) * noise
# None
if rec_type == 0:
p = mem_unit.infer_max(img)[0]
elif rec_type == 1:
p = mem_unit.infer_argmax(img)[0]
mse += torch.mean(torch.square(img.view(img.size(0), -1) - p.view(img.size(0), -1)), dim=1).sum().item()
recalled += ((torch.mean(torch.square(img.view(img.size(0), -1) - p.view(p.size(0), -1)),
dim=1) <= rec_thr).sum()).item()
mse /= mem_images.size(0)
recalled /= mem_images.size(0)
return mse, recalled
def get_train_energies(model, mem_images, engs, seed, dev='cuda'):
with torch.no_grad():
images = mem_images.clone()
#Just FF
a = model.ff_max(images)
engs[0][seed,0] = model.compute_energy(images, a)
a = model.ff_argmax(images)
engs[1][seed, 0] = model.compute_energy(images, a)
a = model.ff_softmax(images, beta=100)
engs[2][seed, 0] = model.compute_energy(images, a)
a = [torch.rand_like(a[x]) for x in range(len(a))]
engs[3][seed, 0] = model.compute_energy(images, a)
# FF + FB
a = model.infer_max(images)
engs[0][seed, 1] = model.compute_energy(images, a)
a = model.infer_argmax(images)
engs[1][seed, 1] = model.compute_energy(images, a)
a = model.infer_softmax(images, beta=100)
engs[2][seed, 1] = model.compute_energy(images, a)
a = [torch.rand_like(a[x]) for x in range(len(a))]
engs[3][seed, 1] = model.compute_energy(images, a)
# 3x FF + FB
a = model.infer_max_iter(images)
engs[0][seed, 2] = model.compute_energy(images, a)
def get_test_energies(model, test_loader, engs, seed, dev='cuda'):
with torch.no_grad():
for batch_idx, (images, y) in enumerate(test_loader):
images = images[0:1000].to(dev)
#Just FF
a = model.ff_max(images)
engs[0][seed,0] = model.compute_energy(images, a)
a = model.ff_argmax(images)
engs[1][seed, 0] = model.compute_energy(images, a)
a = model.ff_softmax(images, beta=100)
engs[2][seed, 0] = model.compute_energy(images, a)
a = [torch.rand_like(a[x]) for x in range(len(a))]
engs[3][seed, 0] = model.compute_energy(images, a)
# FF + FB
a = model.infer_max(images)
engs[0][seed, 1] = model.compute_energy(images, a)
a = model.infer_argmax(images)
engs[1][seed, 1] = model.compute_energy(images, a)
a = model.infer_softmax(images, beta=100)
engs[2][seed, 1] = model.compute_energy(images, a)
a = [torch.rand_like(a[x]) for x in range(len(a))]
engs[3][seed, 1] = model.compute_energy(images, a)
#3x FF + FB
a = model.infer_max_iter(images)
engs[0][seed, 2] = model.compute_energy(images, a)
break
#train_online(arch=2, data=4, dev='cuda', max_iter=200, wtupType=0, num_seeds=5, alpha=10000000, in_dim=64, in_chn=3, shuf=True)
train_online(arch=2, data=4, dev='cuda', max_iter=1000, wtupType=0, num_seeds=3, alpha=10000000, in_dim=64, in_chn=3, shuf=True)