In [1]:
import torch
import torch.nn as nn
import torchvision
import torchvision.transforms as transforms
In [2]:
# Device configuration
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
In [3]:
# Hyper-parameters
input_size = 784
hidden_size = 500
num_classes = 10
num_epochs = 5
batch_size = 100
learning_rate = 0.001
In [4]:
# MNIST dataset
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 [9]:
# Data loader
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 [11]:
# Fully connected neural network with one hidden layer
class NeuralNet(nn.Module):
def __init__(self, input_size, hidden_size, num_classes):
super(NeuralNet, self).__init__()
self.fc1 = nn.Linear(input_size, hidden_size)
self.relu = nn.ReLU()
self.fc2 = nn.Linear(hidden_size, num_classes)
def forward(self, x):
out = self.fc1(x)
out = self.relu(out)
out = self.fc2(out)
return out
In [13]:
model = NeuralNet(input_size, hidden_size, num_classes).to(device)
In [15]:
# Loss and optimizer
criterion = nn.CrossEntropyLoss()
optimizer = torch.optim.Adam(model.parameters(), lr=learning_rate)
In [17]:
# Train the model
total_step = len(train_loader)
for epoch in range(num_epochs):
for i, (images, labels) in enumerate(train_loader):
# Move tensors to the configured device
images = images.reshape(-1, 28*28).to(device)
labels = labels.to(device)
# 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: 0.2807 Epoch [1/5], Step [200/600], Loss: 0.2587 Epoch [1/5], Step [300/600], Loss: 0.1431 Epoch [1/5], Step [400/600], Loss: 0.1042 Epoch [1/5], Step [500/600], Loss: 0.1094 Epoch [1/5], Step [600/600], Loss: 0.0942 Epoch [2/5], Step [100/600], Loss: 0.1129 Epoch [2/5], Step [200/600], Loss: 0.1657 Epoch [2/5], Step [300/600], Loss: 0.0813 Epoch [2/5], Step [400/600], Loss: 0.1026 Epoch [2/5], Step [500/600], Loss: 0.0576 Epoch [2/5], Step [600/600], Loss: 0.1254 Epoch [3/5], Step [100/600], Loss: 0.0597 Epoch [3/5], Step [200/600], Loss: 0.0452 Epoch [3/5], Step [300/600], Loss: 0.0943 Epoch [3/5], Step [400/600], Loss: 0.0997 Epoch [3/5], Step [500/600], Loss: 0.0445 Epoch [3/5], Step [600/600], Loss: 0.0866 Epoch [4/5], Step [100/600], Loss: 0.0914 Epoch [4/5], Step [200/600], Loss: 0.0646 Epoch [4/5], Step [300/600], Loss: 0.0601 Epoch [4/5], Step [400/600], Loss: 0.0155 Epoch [4/5], Step [500/600], Loss: 0.0455 Epoch [4/5], Step [600/600], Loss: 0.0547 Epoch [5/5], Step [100/600], Loss: 0.0772 Epoch [5/5], Step [200/600], Loss: 0.0442 Epoch [5/5], Step [300/600], Loss: 0.0329 Epoch [5/5], Step [400/600], Loss: 0.0339 Epoch [5/5], Step [500/600], Loss: 0.0457 Epoch [5/5], Step [600/600], Loss: 0.0323
In [23]:
# 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, 28*28).to(device)
labels = labels.to(device)
outputs = model(images)
_, predicted = torch.max(outputs.data, 1)
total += labels.size(0)
correct += (predicted == labels).sum().item()
print('Accuracy of the network on the 10000 test images: {} %'.format(100 * correct / total))
Accuracy of the network on the 10000 test images: 97.86 %
In [25]:
# Save the model checkpoint
torch.save(model.state_dict(), 'model.ckpt')
In [ ]: