파이토치로 nn모듈의 CNN사용하기

파이토치로 CNN 모델 생성(fasion_mnist data)

from torch.utils.data import Dataset, DataLoader
import torchvision.transforms as transforms
from torchvision import datasets
import torch.nn as nn
import torch.nn.functional as F
import torch
import torch.optim as optim

#ToTensor() - 데이터를 tensor로 바꿔준다.
#Normalize(mean, std, inplace=False) - 정규화한다.
# Fasion-mnist 데이터 불러오기
transform = transforms.Compose([transforms.ToTensor()])
trainset = datasets.FashionMNIST(root='/content',
                                 train=True, download=True,
                                 transform = transform)
testset = datasets.FashionMNIST(root='/content',
                                 train=False, download=True,
                                 transform = transform)

#데이터로더로 데이터 정제
train_loader = DataLoader(trainset, batch_size=64, shuffle=True, num_workers=2)
test_loader = DataLoader(testset, batch_size=64, shuffle=False, num_workers=2)

#반복문에서 사용할 수 있도록 iter로 만들기
images, labels = next(iter(train_loader))
images.shape, labels.shape #(torch.Size([64, 1, 28, 28]), torch.Size([64]))

#nn.Module 클래스 상속(CNN 모델 만들기)
class FashionCNN(nn.Module):
    
    def __init__(self):
        super(FashionCNN, self).__init__()

        #    ImgIn shape=(1, 28, 28, 1)
        #    Conv     -> (1, 28, 28, 32)
        #    Pool     -> (1, 14, 14, 32) 
        self.layer1 = nn.Sequential(
            nn.Conv2d(1,32,3, padding=1), #padding을 하여 이미지 크기 손실이 없음
            nn.BatchNorm2d(32),
            nn.ReLU(),
            nn.MaxPool2d(2,2)
        )

        #    ImgIn shape=(1, 14, 14, 1)
        #    Conv     -> (1, 12, 12, 64)
        #    Pool     -> (1, 6, 6, 64)
        self.layer2 = nn.Sequential(
            nn.Conv2d(32,64,3), #padding옵션이 없어서 이미지 크기가 -2되었음(kenel size=3이기때문에)
            nn.BatchNorm2d(64),
            nn.ReLU(),
            nn.MaxPool2d(2)
        )
        
        self.fc1 = nn.Linear(64*6*6, 600) #layer2출력이 64, 이미지 사이즈가 6*6이기 때문에 1차원으로 표현하면 64*6*6
        self.drop = nn.Dropout2d(0.25)
        self.fc2 = nn.Linear(600, 120)
        self.fc3 = nn.Linear(120, 10)
        
    def forward(self, x):
        out = self.layer1(x)
        out = self.layer2(out)
        out = out.view(out.size(0), -1) #64*6*6(1차원으로 변경)
        out = self.fc1(out)
        out = self.drop(out)
        out = self.fc2(out)
        out = self.fc3(out)
        
        return out
model = FashionCNN()

#손실함수, 옵티마이저 정의
criterion = nn.CrossEntropyLoss()
optimizer = optim.SGD(model.parameters(), lr=0.001,momentum=0.9)

#학습 (20epochs)
for epoch in range(20):
  running_loss = 0.0

  for i, data in enumerate(train_loader, 0):
    inputs , labels = data

    optimizer.zero_grad()

    outputs = model(inputs)
    loss = criterion(outputs, labels)
    loss.backward()
    optimizer.step()

    running_loss += loss.item()
    #item() == tensor에 저장된 값만 가져옴(텐서에서 값만 가져옴)

    if i % 100 == 99:
      print('Epoch : {}, Iter : {}, Loss : {}'.format(epoch+1,i+1, running_loss/2000))
      runngin_loss = 0.0

#평가
correct = 0
total = 0
with torch.no_grad():
  for data in test_loader:
    images, labels = data
    outputs = model(images)
    _, predicted = torch.max(outputs.data, 1)
    total += labels.size(0)
    correct += (predicted == labels).sum().item()

print(100 * correct / total)
#90.69

#저장
PATH = './fashion_mnist.pth'
torch.save(model.state_dict(), PATH)
#불러오기
model = FashionCNN()
model.load_state_dict(torch.load(PATH))