Logistic_regression

Pytorch实现Logistic_Regression

# Import necessary packages.
import torch
import torch.nn as nn
import torchvision
import torchvision.transforms as transforms

# Set the Hyper-parameters.
input_size = 28 * 28    # 784
num_classes = 10
num_epochs = 10
batch_size = 100
learning_rate = 0.001

# Load MINST 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())

# Define the Data Loader (input pipline).
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)

# Define the Logistic regression model.
model = nn.Linear(input_size, num_classes)

# Define the Loss and Optimizer.
# We use the nn.CrossEntropyLoss() computes softmax internally
criterion = nn.CrossEntropyLoss()
optimizer = torch.optim.SGD(model.parameters(), lr=learning_rate)

# 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()

        # Set an output counter.
        if (i+1) % 100 ==0:
            print('Epoch [{}/{}], Step [{}/{}], Loss:{:.4f}'
                  .format(epoch+1, num_epochs, i+1, total_step, loss.item()))

Epoch [1/10], Step [100/600], Loss:2.2729
Epoch [1/10], Step [200/600], Loss:2.2154
Epoch [1/10], Step [300/600], Loss:2.0251
Epoch [1/10], Step [400/600], Loss:1.9289
Epoch [1/10], Step [500/600], Loss:1.8836
Epoch [1/10], Step [600/600], Loss:1.8227
Epoch [2/10], Step [100/600], Loss:1.7710
Epoch [2/10], Step [200/600], Loss:1.6861
Epoch [2/10], Step [300/600], Loss:1.7019
Epoch [2/10], Step [400/600], Loss:1.5922
Epoch [2/10], Step [500/600], Loss:1.5029
Epoch [2/10], Step [600/600], Loss:1.5454
Epoch [3/10], Step [100/600], Loss:1.5381
Epoch [3/10], Step [200/600], Loss:1.3151
Epoch [3/10], Step [300/600], Loss:1.4771
Epoch [3/10], Step [400/600], Loss:1.2503
Epoch [3/10], Step [500/600], Loss:1.3210
Epoch [3/10], Step [600/600], Loss:1.3356
Epoch [4/10], Step [100/600], Loss:1.2238
Epoch [4/10], Step [200/600], Loss:1.1496
Epoch [4/10], Step [300/600], Loss:1.2525
Epoch [4/10], Step [400/600], Loss:1.1648
Epoch [4/10], Step [500/600], Loss:1.0497
Epoch [4/10], Step [600/600], Loss:1.1387
Epoch [5/10], Step [100/600], Loss:1.0765
Epoch [5/10], Step [200/600], Loss:1.0479
Epoch [5/10], Step [300/600], Loss:1.0902
Epoch [5/10], Step [400/600], Loss:1.0168
Epoch [5/10], Step [500/600], Loss:0.9326
Epoch [5/10], Step [600/600], Loss:1.0073
Epoch [6/10], Step [100/600], Loss:1.0087
Epoch [6/10], Step [200/600], Loss:0.9378
Epoch [6/10], Step [300/600], Loss:0.9467
Epoch [6/10], Step [400/600], Loss:0.9626
Epoch [6/10], Step [500/600], Loss:1.0194
Epoch [6/10], Step [600/600], Loss:0.8813
Epoch [7/10], Step [100/600], Loss:0.9351
Epoch [7/10], Step [200/600], Loss:0.8565
Epoch [7/10], Step [300/600], Loss:0.8663
Epoch [7/10], Step [400/600], Loss:0.8950
Epoch [7/10], Step [500/600], Loss:0.9888
Epoch [7/10], Step [600/600], Loss:0.7603
Epoch [8/10], Step [100/600], Loss:0.7935
Epoch [8/10], Step [200/600], Loss:0.8608
Epoch [8/10], Step [300/600], Loss:0.9538
Epoch [8/10], Step [400/600], Loss:0.7843
Epoch [8/10], Step [500/600], Loss:0.7662
Epoch [8/10], Step [600/600], Loss:0.8472
Epoch [9/10], Step [100/600], Loss:0.8384
Epoch [9/10], Step [200/600], Loss:0.8758
Epoch [9/10], Step [300/600], Loss:0.7508
Epoch [9/10], Step [400/600], Loss:0.7223
Epoch [9/10], Step [500/600], Loss:0.7314
Epoch [9/10], Step [600/600], Loss:0.6978
Epoch [10/10], Step [100/600], Loss:0.6625
Epoch [10/10], Step [200/600], Loss:0.7967
Epoch [10/10], Step [300/600], Loss:0.7854
Epoch [10/10], Step [400/600], Loss:0.6818
Epoch [10/10], Step [500/600], Loss:0.7990
Epoch [10/10], Step [600/600], Loss:0.5855

# Test the model.
# In test phase, we don't need to comput gradients (for memory efficiency)
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: 85.55999755859375 %

# Save the model checkpoint.
torch.save(model.state_dict(), 'model_param.ckpt')
# torch.save(model, 'model.ckpt')

# Load the model checkpoint.
model = model.load_state_dict(torch.load('model_param.ckpt'))