In [1]:
# Import packages
import torch
import torch.nn as nn
import torchvision
import torchvision.transforms as transforms
In [3]:
# Hyper-parameters
input_size = 28 * 28 # 784
num_classes = 10
num_epochs = 5
batch_size = 100
learning_rate = 0.001
In [5]:
# MNIST dataset (images and labels)
train_dataset = torchvision.datasets.MNIST(root='../../data',
train=True,
transform=transforms.ToTensor(),
download=True)
test_dataset = torchvision.datasets.MNIST(root='../../data',
train=False,
transform=transforms.ToTensor())
In [7]:
# Data loader (input pipeline)
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)
In [9]:
# Define Logistic regression model
model = nn.Linear(input_size, num_classes)
In [11]:
# Loss and optimizer
# nn.CrossEntropyLoss() computes softmax internally
criterion = nn.CrossEntropyLoss()
optimizer = torch.optim.SGD(model.parameters(), lr=learning_rate)
In [13]:
# Train the model
total_step = len(train_loader)
for epoch in range(num_epochs):
for i, (images, labels) in enumerate(train_loader):
# Reshape images to (batch_size, input_size)
images = images.reshape(-1, input_size)
# Forward pass
outputs = model(images)
loss = criterion(outputs, labels)
# Backward and optimize
optimizer.zero_grad()
loss.backward()
optimizer.step()
if (i+1) % 100 == 0:
print ('Epoch [{}/{}], Step [{}/{}], Loss: {:.4f}' .format(epoch+1, num_epochs, i+1, total_step, loss.item()))
Epoch [1/5], Step [100/600], Loss: 2.1880 Epoch [1/5], Step [200/600], Loss: 2.1291 Epoch [1/5], Step [300/600], Loss: 2.0113 Epoch [1/5], Step [400/600], Loss: 1.9134 Epoch [1/5], Step [500/600], Loss: 1.8158 Epoch [1/5], Step [600/600], Loss: 1.7324 Epoch [2/5], Step [100/600], Loss: 1.6793 Epoch [2/5], Step [200/600], Loss: 1.6874 Epoch [2/5], Step [300/600], Loss: 1.6384 Epoch [2/5], Step [400/600], Loss: 1.5392 Epoch [2/5], Step [500/600], Loss: 1.5413 Epoch [2/5], Step [600/600], Loss: 1.4399 Epoch [3/5], Step [100/600], Loss: 1.4636 Epoch [3/5], Step [200/600], Loss: 1.4613 Epoch [3/5], Step [300/600], Loss: 1.3668 Epoch [3/5], Step [400/600], Loss: 1.2384 Epoch [3/5], Step [500/600], Loss: 1.3049 Epoch [3/5], Step [600/600], Loss: 1.2016 Epoch [4/5], Step [100/600], Loss: 1.2799 Epoch [4/5], Step [200/600], Loss: 1.1369 Epoch [4/5], Step [300/600], Loss: 1.1987 Epoch [4/5], Step [400/600], Loss: 1.1251 Epoch [4/5], Step [500/600], Loss: 1.1797 Epoch [4/5], Step [600/600], Loss: 1.1026 Epoch [5/5], Step [100/600], Loss: 1.1037 Epoch [5/5], Step [200/600], Loss: 0.9322 Epoch [5/5], Step [300/600], Loss: 1.1390 Epoch [5/5], Step [400/600], Loss: 1.0940 Epoch [5/5], Step [500/600], Loss: 1.0617 Epoch [5/5], Step [600/600], Loss: 0.9842
In [15]:
# Test the model
# In test phase, we don't need to compute gradients
with torch.no_grad():
correct = 0
total = 0
for images, labels in test_loader:
images = images.reshape(-1, input_size)
outputs = model(images)
_, predicted = torch.max(outputs.data, 1)
total += labels.size(0)
correct += (predicted == labels).sum()
print('Accuracy of the model on the 10000 test images: {} %'.format(100 * correct / total))
Accuracy of the model on the 10000 test images: 82.62000274658203 %
In [17]:
# Save the model checkpoint
torch.save(model.state_dict(), 'model.ckpt')