p5:基于pytorch的Linear Regression

import torch
import torch.nn as nn

# 1.准备数据集
x_data = torch.Tensor([[1.0], [2.0], [3.0]])  # 1维的，1个特征，3*1，  必须是tensor
y_data = torch.Tensor([[2.0], [4.0], [6.0]])   # 输出1维. 1个特征，3*1


# 2.定义模型
class LinearModel(nn.Module):
    def __init__(self):  #####
        super(LinearModel, self).__init__()  #####
        self.linear = nn.Linear(1, 1)  # (1,1)表示x,y都只有一个特征(行表示样本数，列表示特征)

    def forward(self, x):
        y_pred = self.linear(x)  # linear为callable,将x传入具体运用
        return y_pred


model = LinearModel()  # creat a instance of class LinearModel，model is callable, using model(x)
# 3.构造损失函数和优化器
loss_fn = nn.MSELoss(size_average=False)  # True相当于除以N/1 criterion需要y_hat和y则可求出loss
optimizer = torch.optim.SGD(model.parameters(),
                            lr=0.01)  # 优化器不会构建计算图，不继承Module model.parameters()表优化器对哪些权重做优化和学习率，将来直接用其对模型优化
# 4. training cycle
for epoch in range(1000):
    # forward
    y_pred = model(x_data)
    loss = loss_fn(y_pred, y_data)
    print(epoch, loss.item())

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

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

# Test Model
x_test = torch.Tensor([[4.0]])  # 输入是1*1的矩阵
y_test = model(x_test)  # 输出是1*1的矩阵
print('y_pred = ', y_test.data)

调用关系：model.linear.weight.item()用来拿到w和b

epoch=100

epoch=1000

参考视频：https://www.bilibili.com/video/BV1Y7411d7Ys?p=5