pytorch 05 pytorch实现线性回归

import torch

x_data = torch.Tensor([[1], [2], [3]])
y_data = torch.Tensor([[4], [5], [6]])

class LinearModel(torch.nn.Module):
    def __init__(self):
        super(LinearModel,self).__init__()  # 继承父类的初始化函数
        self.linear = torch.nn.Linear(1,1)  # 对输入的数据做线性变换


    def forward(self,x):
        y_pred = self.linear(x)             # 前向传播
        return  y_pred


model = LinearModel()              # 实例化对象

criterion = torch.nn.MSELoss(size_average=False)         # 定义损失函数
optimizer = torch.optim.SGD(model.parameters(), lr=0.01)           # 定义优化器

for epoch in range(100):               # 训练过程
    y_pred = model(x_data)        # 计算预测值
    loss = criterion(y_pred,y_data)    # 计算损失函数
    print(epoch, loss)          # 打印当前轮次和损失值

    optimizer.zero_grad()        # 梯度归零
    loss.backward()              # 反向传播
    optimizer.step()             # 更新


print('w=', model.linear.weight.item())     # 打印权重   加了item()不显示矩阵，只显示数值，返回标量值
print('b=', model.linear.bias.item())       # 打印偏置

# Test Model
x_test = torch.Tensor([[4]])
y_test = model(x_test)
print('y_pred = ', y_test.data)

运行结果

......................
96 tensor(0.4340, grad_fn=<MseLossBackward0>)
97 tensor(0.4278, grad_fn=<MseLossBackward0>)
98 tensor(0.4217, grad_fn=<MseLossBackward0>)
99 tensor(0.4156, grad_fn=<MseLossBackward0>)
w= 1.4291688203811646
b= 2.0243985652923584
y_pred =  tensor([[7.7411]])

只循环了100次，在测试集上表现不算特别好

作业代码示例，通过绘图来观察

import torch
import matplotlib.pyplot as plt

x_data = torch.Tensor([[1], [2], [3]])
y_data = torch.Tensor([[4], [5], [6]])

class LinearModel(torch.nn.Module):
    def __init__(self):
        super(LinearModel,self).__init__()  # 继承父类的初始化函数
        self.linear = torch.nn.Linear(1,1)  # 对输入的数据做线性变换


    def forward(self,x):
        y_pred = self.linear(x)             # 前向传播
        return  y_pred


model = LinearModel()              # 实例化对象

criterion = torch.nn.MSELoss(size_average=False)         # 定义损失函数
optimizer = torch.optim.ASGD(model.parameters(), lr=0.01)       # 定义优化器,pytorch有很多种优化器可以换

losses = []
epoches = []

for epoch in range(100):               # 训练过程
    y_pred = model(x_data)        # 计算预测值
    loss = criterion(y_pred,y_data)    # 计算损失函数
    losses.append(loss.item())
    epoches.append(epoch)
    print(epoch, loss)          # 打印当前轮次和损失值

    optimizer.zero_grad()        # 梯度归零
    loss.backward()              # 反向传播
    optimizer.step()             # 更新


print('w=', model.linear.weight.item())     # 打印权重   加了item()不显示矩阵，只显示数值，返回标量值
print('b=', model.linear.bias.item())       # 打印偏置

# Test Model
x_test = torch.Tensor([[4]])
y_test = model(x_test)
print('y_pred = ', y_test.data)

plt.plot(epoches, losses)
plt.xlabel('Epochs')
plt.ylabel('Loss')
plt.title('Training Loss vs. Epochs')
plt.show()