FashionMNIST 데이터셋으로 기본적인 오코인코더 모델 구현
|
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 numpy as np
import os
import matplotlib.pyplot as plt
import torch.nn as nn
import torch.nn.functional as F
from torchvision import transforms, datasets
import torch.nn.init as init
os.environ['KMP_DUPLICATE_LIB_OK'] = 'True'
BATCH_SIZE = 64
EPOCHS = 10
if torch.cuda.is_available():
DEVICE = torch.device('cuda')
else:
DEVICE = torch.device('cpu')
print(DEVICE)
train_dataset = datasets.FashionMNIST(root="./data/FashionMNIST",
train=True,
download=True,
transform=transforms.ToTensor())
test_dataset = datasets.FashionMNIST(root="./data/FashionMNIST",
train=False,
download=True,
transform=transforms.ToTensor())
train_loader = torch.utils.data.DataLoader(dataset=train_dataset,
batch_size=BATCH_SIZE,
shuffle=True)
test_loader = torch.utils.data.DataLoader(dataset=test_dataset,
batch_size=BATCH_SIZE,
shuffle=False)
# 다운로드 받은 데이터셋 확인
for (x_train, y_train) in train_loader:
print('x_train: ', x_train.size(), ' data_type: ', x_train.type())
print('y_train: ', y_train.size(), ' data_type: ', y_train.type())
break
fig = plt.figure(figsize=(5, 1))
for i in range(5):
plt.subplot(1, 5, i + 1)
plt.axis('off')
plt.imshow(x_train[i, :, :, :].numpy().reshape(28, 28), cmap="gray_r")
plt.title("class: " + str(y_train[i].item()))
plt.show()
class AutoEncoder(nn.Module):
def __init__(self):
super(AutoEncoder, self).__init__()
self.encoder = nn.Sequential(
nn.Linear(28 * 28, 512),
nn.ReLU(),
nn.Linear(512, 256),
nn.ReLU(),
nn.Linear(256, 32),
)
self.decoder = nn.Sequential(
nn.Linear(32, 256),
nn.ReLU(),
nn.Linear(256, 512),
nn.ReLU(),
nn.Linear(512, 28*28),
)
def forward(self, x):
encoding = self.encoder(x)
decoding = self.decoder(encoding)
return encoding, decoding
def weight_initializer(m):
if isinstance(m, nn.Linear):
init.kaiming_uniform_(m.weight.data)
model = AutoEncoder().to(DEVICE)
model.apply(weight_initializer) #가중치 초기화 기법을 사용
optimizer = torch.optim.Adam(model.parameters(), lr=1e-3)
criterion = nn.MSELoss()
print(model)
def train(model, train_loader, optimizer, interval):
model.train()
for idx, (image, _) in enumerate(train_loader):
image = image.view(-1, 28*28).to(DEVICE)
target = image.view(-1, 28*28).to(DEVICE)
optimizer.zero_grad()
encoding, decoding = model(image)
loss = criterion(decoding, target)
loss.backward()
optimizer.step()
if idx % interval == 0:
print('train epoch: {}, {}/{} train_loss: {}'
.format(epoch, idx*len(image), len(train_loader.dataset), loss.item()))
def evaluate(model, test_loader):
model.eval()
test_loss = 0
input_image = []
reconstruct_image = []
with torch.no_grad():
for image, _ in test_loader:
image = image.view(-1, 28*28).to(DEVICE)
target = image.view(-1, 28*28).to(DEVICE)
encoding, decoding = model(image)
test_loss += criterion(decoding, image).item()
input_image.append(image.to("cpu"))
reconstruct_image.append(decoding.to("cpu"))
test_loss /= len(test_loader.dataset)
return test_loss, input_image, reconstruct_image
for epoch in range(1, EPOCHS+1):
train(model, train_loader, optimizer, 200)
test_loss, input_image, reconstruct_image = evaluate(model, test_loader)
print("test_loss: {}".format(test_loss))
_, a = plt.subplots(2, 10, figsize=(10, 4))
for i in range(10):
temp = np.reshape(input_image[0][i], (28, 28))
a[0][i].imshow(temp, cmap="gray_r")
a[0][i].set_xticks(())
a[0][i].set_yticks(())
for i in range(10):
temp = np.reshape(reconstruct_image[0][i], (28, 28))
a[1][i].imshow(temp, cmap="gray_r")
a[1][i].set_xticks(())
a[1][i].set_yticks(())
if epoch == EPOCHS:
plt.show()
|
cs |
위의 소스코드를 실행시키면, 아래와 같은 결과를 얻을 수 있다.
반응형
'머신러닝_딥러닝 > Pytorch' 카테고리의 다른 글
| CNN 1탄 (CIFAR-10 dataset) (0) | 2021.10.17 |
|---|---|
| MLP모델 (CIFAR-10 dataset) (0) | 2021.10.17 |
| 오토인코더 (AutoEncoder) (0) | 2021.10.17 |
| MLP모델 설계 및 학습 2탄 (MNIST dataset) (0) | 2021.10.16 |
| MLP모델 설계 및 학습 1탄 (MNIST dataset) (0) | 2021.10.16 |