1
2
3
4
import torch
import torchvision.datasets as datasets
import torchvision.transforms as transforms
import random
1
2
3
4
5
6
7
device = 'cuda' if torch.cuda.is_available() else 'cpu'
# seed 고정
random.seed(111)
torch.manual_seed(777)
if device == 'cuda':
torch.cuda.manual_seed_all(777)
1
2
3
4
# parameters
learning_rate = 0.001
epochs = 15
batch_size = 100
1
2
3
4
5
6
7
mnist_train = datasets.MNIST(root='./', train=True,
transform=transforms.ToTensor(),
download=True)
mnist_test = datasets.MNIST(root='./', train=False,
transform=transforms.ToTensor(),
download=True)
1
2
3
4
5
# Dataset Loader
data_loader = torch.utils.data.DataLoader(dataset=mnist_train,
batch_size=batch_size,
shuffle=True,
drop_last=True)
Activation Function
torch.nn.relu: ReLUtorch.nn.sigmoid: Sigmoidtorch.nn.Leaky_relu: Leaky ReLU
1
2
3
4
5
6
7
8
# 784인 이유는 28 x 28
# nn.Linear(input_dim, output_dim)
linear1 = torch.nn.Linear(784, 256, bias=True)
linear2 = torch.nn.Linear(256, 256, bias=True)
linear3 = torch.nn.Linear(256, 10, bias=True)
# activation function : ReLU
relu = torch.nn.ReLU()
가중치 초기화
torch.nn.init.normal_: 기본적인 방법torch.nn.init.xavier_uniform_: Xvier
1
2
3
4
# Initialization
torch.nn.init.normal_(linear1.weight)
torch.nn.init.normal_(linear2.weight)
torch.nn.init.normal_(linear3.weight)
1
2
3
4
5
6
7
8
9
Parameter containing:
tensor([[-2.3518, -1.0384, 0.3364, ..., -0.0419, -2.8471, -0.7893],
[-1.3143, 0.4450, 0.0218, ..., -0.1175, 0.4807, -0.0046],
[-1.0058, -1.2789, 0.4103, ..., -0.0894, -0.4732, -1.7324],
...,
[-0.3238, -1.3617, 0.9380, ..., -0.4760, -0.0939, -0.4173],
[ 0.4770, -0.8255, 1.8754, ..., 1.0181, -0.2816, -1.3303],
[ 0.8517, 0.9480, 0.4240, ..., 0.5320, 0.6457, -1.3825]],
requires_grad=True)
1
2
# model
model = torch.nn.Sequential(linear1, relu, linear2, relu, linear3).to(device)
Loss & Optimizer
Loss
torch.nn.BCELoss: 이진 분류torch.nn.CrossEntropyLoss: 다중 클래스 분류torch.nn.MSELoss: MSE
Optimizer
torch.optim.Adam: Adamtorch.optim.SGD: SGDtorch.optim.RMSprop: RMSproptorch.optim.Rprop: Rprop
1
2
3
# loss & optimizer
criterion = torch.nn.CrossEntropyLoss().to(device)
optimizer = torch.optim.Adam(model.parameters(), lr=learning_rate)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
total_batch = len(data_loader)
for epoch in range(epochs):
avg_cost = 0
for X, y in data_loader:
X = X.view(-1, 28 * 28).to(device)
y = y.to(device)
optimizer.zero_grad()
hypothesis = model(X)
cost = criterion(hypothesis, y)
cost.backward()
optimizer.step()
avg_cost += cost / total_batch
print('Epoch: {} Cost: {:.4f}'.format(epoch + 1, avg_cost))
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
Epoch: 1 Cost: 164.3944
Epoch: 2 Cost: 38.7940
Epoch: 3 Cost: 24.0848
Epoch: 4 Cost: 16.4266
Epoch: 5 Cost: 11.6766
Epoch: 6 Cost: 8.5664
Epoch: 7 Cost: 6.3255
Epoch: 8 Cost: 4.5684
Epoch: 9 Cost: 3.4953
Epoch: 10 Cost: 2.5707
Epoch: 11 Cost: 1.9859
Epoch: 12 Cost: 1.4931
Epoch: 13 Cost: 1.1411
Epoch: 14 Cost: 0.8788
Epoch: 15 Cost: 0.6823
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
with torch.no_grad():
X_test = mnist_test.test_data.view(-1, 28 * 28).float().to(device)
y_test = mnist_test.test_labels.to(device)
predict = model(X_test)
correct = torch.argmax(predict, 1) == y_test
accuracy = correct.float().mean()
print('Accuracy: {}'.format(accuracy.item()))
# 하나의 예측
r = random.randint(0, len(mnist_test) - 1)
X_data = mnist_test.test_data[r:r + 1].view(-1, 28 * 28).float().to(device)
y_data = mnist_test.test_labels[r:r + 1].to(device)
print('Label: {}'.format(y_data.item()))
single_predict = model(X_data)
print('Predict: {}'.format(torch.argmax(single_predict, 1).item()))
1
2
3
4
5
6
7
8
9
c:\Users\USER\anaconda3\lib\site-packages\torchvision\datasets\mnist.py:80: UserWarning: test_data has been renamed data
warnings.warn("test_data has been renamed data")
c:\Users\USER\anaconda3\lib\site-packages\torchvision\datasets\mnist.py:70: UserWarning: test_labels has been renamed targets
warnings.warn("test_labels has been renamed targets")
Accuracy: 0.9441999793052673
Label: 8
Predict: 8