04_Linear_Regression(线性回归)

Linear_Regression with pytorch

import torch

x_data = torch.Tensor([[1.0],[2.0],[3.0]])
y_data = torch.Tensor([[2.0],[4.0],[6.0]])

class LinearModel(torch.nn.Module):
    def __init__(self):
        super(LinearModel,self).__init__()
        self.linear = torch.nn.Linear(1,1)


    def forward(self,x):
        y_pred = self.linear(x)
        return y_pred
model = LinearModel()

criterion = torch.nn.MSELoss(size_average=False)
optimizer = torch.optim.SGD(model.parameters(),lr=0.01)

for epoch in range(1000):
    y_pred=model(x_data)
    loss=criterion(y_pred,y_data)
    print(epoch,loss.item())

    optimizer.zero_grad()
    loss.backward()
    optimizer.step()


print('w=',model.linear.weight.item())
print('b=',model.linear.bias.item())

x_test=torch.Tensor([4.0])
y_test=model(x_test)

print('y_pred=',y_test.data)